GB2129971A - Group supervisory control apparatus for lift system - Google Patents

Abstract

A plurality of characteristic modes represent traffic demand states occurring in a building in which the system is installed. When a detected traffic demand state coincides with one of the modes, operation of the system is controlled with the aid of control parameters determined thereby so that the currently detected traffic demand is processed in an optimal manner. The modes are stored and selectively read out in response to traffic demand detection. A mode creating and modifying capability accommodates new traffic demand states through a learning procedure, a mode being established by an automatic learning procedure in dependence on the results of evaluation of the traffic demand characteristic with regard to the significance, the time or period and continuity thereof. To this end, a disconnectable learning microcomputer is provided in addition to the system operation and car controlling microcomputers. Special modes can be manually input at a terminal. <IMAGE>

Description

SPECIFICATION Group supervisory control apparatus for elevator system The present invention relates to an elevator control and particularly to an elevator system in which control with the aid of computers is adopted in a very convenient and advantageous manner.

In recent years, microcomputers are increasingly employed in various industrial fields. In the field of the elevator technology, the microcomputer tends to be increasingly employed not only in a group supervisory control system for managing a plurality of elevator cars in a group but also in a car control apparatus for controlling the individual car with an attempt to enhance the operation and/or service efficiency. Such attempt has been made significant contribution to the improvement of the elevator control apparatus by virtue of the advantageous features of the microcomputer such as small size, high functional ability, high reliability and inexpensiveness.

In the case of the group supervisory control, for example, it is possible to monitor on-line the hail calls individually and select the optimal cars for assigning them to the hall calls in consideration of the state of service being made for the hall calls on the whole to thereby reduce the waiting time, to a great advantage. Further, a preference service control can be performed in such manner that a number of cars are dispatched to the floor which is congested with a large number of passengers or car service is performed to the floor of directrate with a significantly shortened waiting time. In other words, it is possible to perform fine and scrupulous control with the aid of the microcomputer.

On the other hand, in conjunction with the elevator supervising apparatus, there has been proposed such a significantly sophisticated computerized system in which a system processor for the group supervisory control is connected to a central processor installed at an executive center through telephone line to accomplish the supervisory control with high efficiency. In that case, the system processor may be disconnected from the elevator system in the time zone such as the night-time in which no car operation is required, and connected to a car response simulating apparatus incorporated in the central processor of the executive center to monitor efficiently the functions and operations executed by the system processor (reference is to be made to Japanese Patent Publication No.

37145/1981).

However, the prior art group supervisory control system for the elevators cannot adapt itself to the environmental conditions of a building which vary from time to time, because the group supervisory control system is operated by using predetermined and fixed control functions and parameters. Taking as example a traffic demand in a building, the destination-dependent traffic demand will become different from the state prevailed at the time of the opening of the building, when there are changes in the tenants or residents in the building or in respect to the contents of businesses. Further, the traffic demand in a day will vary significantly in dependence on different time zones such as office-going time zone, lunch time zone, office-leaving time zone and the ordinary time zone.

When the traffic demand varies remarkably in this way, difficulty is encountered in performing the supervisory control with a desired efficiency, to degrade the service capability. Under the circumstance, there has been proposed another elevator control system in which the traffic demand in an elevator system is detected to determine the current traffic pattern by comparing the detected traffic demands with a plurality of predetermined typical characteristic modes of the traffic demand which make appearance characteristically in the particular time zones such as mentioned above, to thereby identify the traffic characteristic mode which approximates most closely to the detected traffic demand, as is disclosed in Japanese Patent Publications Nos. 15502/1973 and 141942/1977.

In other words, in the hitherto known group supervisory control system, the group supervisory control programs are prepared for the prospected traffic demand patterns, respectively, and the program is selected correspondingly in dependence on the change in the traffic demand to effect the control.

However, the above mentioned control method can not accommodate various situations such as alteration from a rental office building to a building occupied by only one company, changes in the tenants in the building, increasing in the number of residents, unexpected situations such as change in the traffic demand due to a newly-founded subway or the like, resulting in that the service provided by the elevator system is degraded. To avoid the inconveniences, re-arrangement or re-construction of the group supervisory control system as well as research on the actual traffic condition are necessarily required, to a disadvantage of the prior art elevator control system.

In the case of a school building and an assembly-hall building, substantially similar traffic demand modes may take place repeatedly, but non-periodically. However, there has been proposed neither means for detecting properly such traffic demand nor means for allowing the elevator operation to be performed so as to conform with the prevailing traffic demand.

A further method has been proposed, according to which the elevator operation is effected in different manners in dependence on weekdays and holidays by utilizing a calendar day signal generated by a calendar day signal generating circuit (reference is to be made to Japanese Laid-Open Patent Application No. 141944/1977).

The above method may be adopted more or less satisfactorily in a single company-only building, provided that the building is not opened every Sunday.

However, there may arise such case that the building is closed from Monday to Friday owing to summer vacation and/or alternative holidays. In that case, the inputting of day-data of holiday and weekday must be correspondingly changed over by means of switches or the like. Further, it can not be determined what kind of measures be provided for Saturday on which a part of members who belong to a trade section or the like attend the office.

An object of the present invention is to provide an elevator control system which is capable of providing services which can accommodate promptly changes in the nature of a building in which the elevator system is installed and changes in environmental or traffic conditions.

According to a general feature of the present invention, an operation program is created for each of the characteristic modes of traffic demand previously created or established. The elevator operation is controlled through the operation program most suited to the current traffic demand.

Above and other objects, features and advantages of the invention will be more apparent when reading the description of preferred embodiments taken in conjunction with the accompanying drawings, in which: Figs. 1 to 8 are views for illustrating the principle of the present invention; Figs. 9 to 11 are circuit diagrams showing an apparatus for creating characteristic modes of traffic demand according to an embodiment of the invention; Figs. 1 2 and 13 are views for illustrating data to be recorded; Figs. 14 to 18 are views for illustrating a hardware construction according to a second embodiment of the invention; Fig. 1 9 is a block diagram illustrating a general composition of software according to a second embodiment of the invention; Figs. 20 to 23 are views for illustrating parameters for simulation;; Fig. 24 is a view showing a table structure used by an operation control microcomputer system; Figs. 25 to 29 illustrate in a flow chart call allotment controls effected by the operation control microcomputer system; Figs. 30 to 37 are views for illustrating structures of tables used by a learning microcomputer system; Figs. 38 to 42 show main flow charts for illustrating softwares for the learning microcomputer system; Fig. 43 is a view for illustrating a method of preparing a destination (floor)-based traffic volume table; Figs. 44 to 48 show flow charts for illustrating execution of simulation; Fig. 49 illustrates in a flow chart an exemplary application of the simulation; Fig. 50 shows in a flow chart a car-number-based simulation; Fig. 51 is a view for illustrating the simulation;; Figs. 52 and 53 are views for illustrating supplementarily the operation of the second embodiment of the invention; Figs. 54 to 56 are views for illustrating improvements of a method of predicting generation of the characteristic anodes; Fig. 57 is a view illustrating an exemplary format for reporting the effects of the leaning system; Fig. 58 is a block diagram showing a version; Fig. 59 is a view illustrating structures of tables used in the system shown in Fig. 58; Fig. 60 is a view illustrating flows of data in the learning system: Fig. 61 is a view for illustrating control modes for the cars in the stand-by state ready for making service to congested floors; Fig. 62 shows a modification of the flow chart shown in Fig. 29; and Fig. 63 shows schematically a general arrangement of a group supervisory control monitor system.

In the following, the present invention will be described in connection with the exemplary embodiments thereof. An apparatus for creating characterizing modes of traffic demand in a lift or elevator system will be described in detail by referring to a concrete example shown in Figs. 9 to 13 and another exemplary embodiment shown in Figs. 14 to 1 8. However, before entering into the elucidation of the exemplary embodiments, the control concept of the present invention will first be described with the aid of Figs. 1 to 8.

Referring to Fig. 1, there is illustrated situations or changes in the traffic demand of an elevator system occurring from 8 a.m. to 2 p.m. or so on a certain day by using element which represent characteristics of the traffic demand, said elevator system being installed in a building having a basement (underground floor) and an overground structure having ten floors to be serviced by the elevator system. It should however be noted that the traffic situation in the daytime is omitted from illustration except for those occurring during the office-going time zone and the lunch time zone. In the figure, a curve C, shows on the time base an element representative of the traffic volume, while time charts or curves 9U to B1 D represent degrees of the traffic congestion or concentration on the floor base in terms of digital values each consisting of three digits. Referring to Fig. 7, a curve C3 shown at the upper portion thereof represents magnitude of a traffic demand characterizing element presenting a particular characteristic feature and produced in the daytime in the same building (details of which will be described hereinafter).

For controlling optimally an elevator system in general and in particular an elevator system which includes a plurality of cars installed and operated in parallel in a group under supervisory control, it would be ideal if prediction could be made as to the starting floors and the destinations of the prospective individual passengers. However, as a matter of fact, a great difficulty will be encountered in having several thousand or more persons register beforehand the respective elevator utilizing schedules, and in reality such registration is utterly impracticable in the building where there are a great number of visitors and/or a lot of events or festivals are held.

Under the circumstances, it is contemplated with the present invention to provide the most effective means for solving the aforementioned problem through the control in which the yesterday's traffic demand is learned and prediction is made as to the today's traffic demand. To this end, the basic principle of the invention resides in that the characteristic modes of the traffic demand in a building under consideration are extracted or sampled, wherein magnitudes of plural elements capable of representing properly the extracted or sampled characteristic modes as well as various elements representing times of occurrence or generation, periods and other factors of the characteristic modes are learned for the purpose of performing the optimum control of the elevator system from the next day on the basis of the predictions.

The traffic demand includes various elements or factors. Among them, important ones imposing great burden on the elevator system are the traffic volume and the inter-floor migration.

A function C(t) representing the traffic volume has heretofore been defined by the following expression (1). Namely, number of passengers carried by elevator per 5 minutes in time zone t C(t)= ( 1 ) number of residents allotted to elevator for service The number of the residents may be considered as the number for criterion and is treated simply as a constant.

The traffic volume function is made use of for comparing magnitude of the traffic demands among the different elevator systems installed in different buildings and permits comparison to a rough degree. More specifically, the value of the traffic volume function C(t) is ordinarily in the range of ca.

4% to ca. 6%, and the traffic volume of the order of 1 2% may reasonably indicate an extremely congested traffic situation. This function is however inadequate for indicating magnitude of the burden or stress imposed on the elevator system. This is mainly because the elevator system is designed in consideration of the number of the residents in the associated building. Further, from the viewpoint of the elevator control, the comparison of the traffic volume between the buildings is rather meaningless.

Therefore, according to the present invention, the value of the traffic volume function C(t) is defined as the number of persons who use the elevator system per unit time which is equal to five minutes.

Next, a traffic volume for a predetermined period (an average traffic volume per day reduced from the traffic volume of more than one day and more preferably of seven or more days which can be determined through the traffic volume learning function having a corresponding time constant) is measured to prepare a frequency distribution chart shown in Fig. 2b. With the aid of this frequency chart, traffic volume evaluation levels ai, .2.... 6 are prepared on the basis of the ratios of times occupied by the traffic demands, respectively, in a predetermined period (e.g. one week) in a building under consideration. These levels are then utilized for determining an evaluation value (traffic volume level function) CV(t) which represents magnitude of an elementary value of the traffic volume C of each traffic demand.

Even for a same traffic volume, there arises significant difference in burden or stress imposed on the elevator system between the situation in which all the floors or landing are uniformly congested and the situation where boarding and alighting passengers are concentrated only on the first (or ground) floor and the second (or first) floor. Accordingly, it is necessary to make it possible to discriminatively recognize these situations in terms of some kind of element.

In consideration of the fact that there has been established no available theory for defining the characteristics of the inter-floor migration (hereinafter also referred to a traffic flow for simplification).

there are adopted the expressions of a congestion-concentrated floor function l(t) and a congestion intensity function V(t) of a given floor in concern relative to all the floors for facilitating the description of the invention. Of course, it is also conceivable to determine a distribution factor for each of the floors to express the traffic situation thereof. It is however impossible to control the elevator service correspondingly for the individual floors with only the distribution factors. Accordingly, it is decided that the floors or landings themselves are included in the element used for indicating the characteristics of the traffic demand in the elevator system under consideration.

In this connection, it should be noted that in consideration of the nature of the elevator control, the congestion-concentrated floor function l(t) (an example of which is given by the expression (2) mentioned below) and the congestion intensity function V(t) (an example of which is given by the expression (3) mentioned below) should preferably be determined for the boarding passengers (designated by a symbol S), the alighting passengers (designated by a symbol R), the travel in the updirection (designated by a symbol U) and travel in the down-direction (designated by a symbol D), respectively, for expressing more appropriately the characteristics of various traffic flows involved in the various traffic demands.

As the elementary functions representing these four types of factors mentioned above, there are conceivable eight evaluation values mentioned below.

(3) Evaluation values indicating factors of distribution of the passengers to be serviced by the elevator in the up-direction: Inu(t)=floor which is n-th (order) in the number of the passengers who get on the elevator in the up direction in the time zone (t-At, t+At) (2) Vu(t)=number of passengers boarding up-cars from the floor Ins(t)/number of passengers boarding up cars in the time zone (t-At, t+At) (3) O Evaluation values indicating factors of distribution of the passengers who get on or board the down-cars: Ing(t); Vng(t).

Q3 Evaluation values indicating factors of distribution of the alighting passengers who get off the cars in the up-direction: InU(t); VnR(t).

Evaluation values indicating factors of distribution of the passengers who get off the cars in the down-direction: InR(t); VnR(t).

In the evaluation values mentioned above, n represent an ordinal variable which is variable within a range of 1 to the maximum number of the floors to be serviced (generally, the maximal number of the floors serviced by the elevator system is from five to forty and most frequently thirteen or so). In the case of the elevator system which is assigned with ten floors to be serviced thereby, it is required to determine as many as eighty evaluation values in total, record and process the values for discrimination.

Under the circumstances, the inventors have attempted to reduce the number of the evaluation values to a range within which the practicability of the evaluation value in representing the characteristics of the traffic flow is maintained, as described below.

In the first place, it is decided that the direction is evaluated in one lot. In other words, analyses are to be performed on the assumption that one and the same floor to be serviced is treated as two separate floors in consideration of the car travelling directions.

According to this concept, it is possible to prepare data representative of the floor on the direction base for the congestion-concentrated floor function l(t) with the effect being unchanged.

As to the magnitude of the ordinal variable n, less importance may be put on the directionimparted landing floors (hereinafter simply called floor, while the floor which is not imparted with the direction factor is referred to as the landing) which provide light load as estimated on the basis of the burden or stress imposed on the elevator control. In other words, it is sufficient from the practical viewpoint to consider several floors which present heavy loads on the elevator control. Accordingly, it is assumed in the following description that the values which the ordinal variable n may take are 1, 2 and 3 in consideration of the convenience of description as well. Thus, the eighty values for evaluation can be reduced down to the twelve. More specifically, the evaluation values for individual elements in concern are defined by the expression (4) to (9) mentioned below.

Isn(t)=floor which is n-th in concentration or congestion of the boarding passengers in the time zone (t-At, t+At) (4) Ws(t)=the number of boarding passengers from the i-th floor in the time zone (t-At, t+At) (5) From the expression (4) and (5), the degree or ratio of congestion Vsn(t) on the floor IS(t) concentrated with the boarding passengers is given as follows::

IRn(fl=floOr which is n-th in concentration of the alighting passengers in the time zone (t-At, t+At) (7) Wq(t)=the number of alighting passengers on the i-th floor in the time zone (t-At, t+At) (8) From the expressions (7) and (8), the degree or ratio of congestion on the floor In(t) which is n-th in the concentration of the alighting passengers is given by

The symbol F represents the number of the floors to be serviced by the elevator system. (In the instant case under consideration, it is assumed that F represents the number of the floors to be serviced in one of the direction.However, this symbol F may represent the number of the floors to be serviced without taking the direction into consideration.) For the purpose of further simplification, the inventors have finally decided to adopt the below mentioned expression (10) and (11) in place of the expressions (6) and (9) representing the congestion intensity function V, to thereby make available the eight evaluation values as the elements which express the characteristics of the traffic flow. Namely,

where Vn represents the degree of concentration (%) on the floor base.

Referring to Fig. 3, there are graphically illustrated typical traffic flows which occur frequently in the office-going time zone (t2-t3) shown in Fig. 1. In Fig. 3, there is illustrated at (a) the degree of concentration or congestion on the floor base, wherein a curve labelled fUPIN1 represents the ratio of the passengers carried by the cars in the up direction In terms of the floor-based distribution while a curve fDNIN1 represents the ratio of concentration of the passengers carried by the cars travelling in the down direction.Curves fUPOUT1 and fDNOUT1 represent distributions of the passengers carried by the cars travelling in the up (ascending) and the down (descending) directions, respectively, wherein the sum of the ratios of the boarding and the alighting passengers amounts to 100%, and the number of the passengers in total is equal to that represented by the traffic volume when considered macroscopically.

Referring to Fig. 3 at (b), there are graphically illustrated the ratios of the passengers on the down-car base in the order of the floors of the highest to the lowest concentration, wherein a curve fVlN represents the degrees of concentration of the boarding passengers at the individual floors while a curve fVOUT represents the degrees of dispersion of the alighting passengers to the individual floors.

From the analyses described above, the traffic demand in the typical office-going time zone It2 t3) illustrated in Figs. 1 and 3 can be reduced to the elements mentioned below through the characteristic recognition procedures. Namely, (1) Traffic volume function C(t), wherein C(trt2A)tt20 men/t min (12) C(t2At3)z200 rnen/t min (13) (2) Floor-based congestion intensity function of the boarding passengers Vs(t), wherein VS(t,--t,)w602+ 302x4 500 (14) (3) Floor-based congestion intensity function af the alighting passengers {VR(t), wherein V1(t2-t3)1 72+162+142+ 133+122+1 12+72+42,,,,,,,1 545 (15) (4) Congestion-concentrated floor function for the boarding passengers Ins(t), wherein Ins(t2t3)=$o1, $02, $oo (16) assuming that values of the ordinal variable n are 1, 2 and 3, respectively.

(5) Congestion-concentrated floor function for the alighting passengers 1R (t), wherein lRn(t2t3)=$O4, $07, $06 (17) assuming that n=1,2 and 3, respectively.

In conjunction with the congestion-concentrated floor functions l(t) given by the expressions (16) and (17), it should be noted that those floors where the ratio of the passengers is smaller than a predetermined value P, shown in Fig. 3 can not be called the congestion-concentrated floor. In consideration of this fact, the floor In of the ordinal number n which is not the congestion-concentrated floor in this sense is allotted with the symbol SOO in the array for indicating that the floor In is not to be considered in the congestion-concentrated floor function.

In this manner, such characteristic aspects of the traffic demand in the office-going hour zone in a predetermined time span (e.g. one day) that the demand is approximately at maximum, the greater part of the boarding passengers are from a particular floor (such as lobby) where ISn=1(fl=$02 (as verified by Vs(t)=4500), and that the passengers travelling in the down direction is extremely small in number (i.e.

$81 to $8B are not available to the function I:(t)) can be expressed by the eight functions taking the five elementary values mentioned above to an utterly satisfactory degree for practical application.

Next, description will be made as to the manner in which the various traffic demands generated in various time zones and the days of the week are recognized and extracted or sampled as the characteristic modes significant for the building under consideration.

In the first place, the general control principle for the elevator system will be elucidated by referring to Fig. 4 which illustrates the control procedures. It should be noted that Fig. 4 illustrate neither programs nor hardware circuits nor flows of operations but conceptional procedures of learning processes. At first, in precedence to the opening day (initiation of elevator operation) of the building under consideration (procedure No. P10), characteristic modes of the prospected traffic demand are inputted with the aid of a keyboard or the like of an intelligent terminal. The data inputted at this time point are of time schedule (table No. T291) and the floors expected to be congested, the floor data being entered in a table No. T292 in the order of the prospected congestions, as is shown in Fig. 32.

Since at least the congestion-concentrated floor is thus known at this step, the five sets of the elementary evaluation values for the eight characteristic aspects in total described hereinbefore are prepared in the manners mentioned below: (1) C(AM 10-AM 10.30)=135 (18) When no designation is made as to the traffic volume, 1 35 (men/t min) corresponding to the middle level a4 is set.

(2) l5(AM 10-AM 10.30)=$82, S05, $86 (19) (3) IR(AM 10-AM 10.30)=$82, 05, $85 (20) In this connection, it should be noted that the same floors are inputted for both 15 and IR since no indication is given as to discrimination between the congestion of the boarding passengers and the alighting passengers. In an alternative case where only the floor data have been inputted, both directions are considered.

(4) V5(AM 10-AM 10.30)=1700 (21) (5) VR(AM 10-AM 10.30)=1700 (22) In this connection, since no indication has been given as to the degree of congestion or concentration and three congestion-expected floors are designated, the value 1700 has been selected which corresponds to the concentration level p2.

Since the characteristic modes such as those occurring in the office-going time zone and the lunch time zone are likely to take place rather universally, they may be placed in a ROM or the like at the time of shipping from the manufacturer's factory. This procedure is, however, required only when the optimal operation of the elevator system is to be realized as of the opening day and is usually unnecessary.

Next, description will be made about the automatic setting of the characteristic modes for the elevator control (procedure P30), Details of this procedure are illustrated in Fig. 5. By way of example, the traffic demand in one day or in one week is detected (procedure P31), thereby to arithmetically determine the five types of the characteristic elements mentioned hereinbefore, i.e. the traffic volume C(t) and the characteristic elementary functions of the traffic flow (procedure P32), which are subsequently recorded (at a step P33). When the detection of the traffic demand is made, for example, every 7.5 minutes, the data obtained in a day will amount to as many as 192 sets which correspond to 1344 sets in a week. Under the circumstance, a nonvolatile memory of about 10 KB capacity will be required for recording the characteristic elementary function values in eight per set. Of course, the leaming operation by extracting the characteristic modes will take correspondingly a lot of time, necessitating arithmetic hardwares operatable at high speed. Besides, difficulty will be encountered in deriving the characteristics when the number of the passengers is small. Accordingly, it is proposed that the traffic demand is detected at a longer period when the traffic volume is small while the traffic demand detection is performed in a shorter period when the traffic volume is great, as is illustrated in Fig. 1, to thereby decrease correspondingly the number of sets of data to be recorded.By way of example, the detection of the characteristic data for the traffic volume of a predetermined number of the passengers will be able to decrease the number of data sets down to ca. 48 per day on the assumption that determination of the characteristics are made every predetermined time, e.g. every 30 min. In this case, it is required to record also the time data indicating the detecting or sampling interval in pair with the sampled characteristic data. The procedure mentioned above is repeated for a predetermined time, e.g. throughout one day (procedure P34), to collect several ten or several hundred sets of the traffic data analyzed for one day. Subsequently, a characteristic mode extracting function is evaluated for determining whether a new characteristic mode is to be established or not.

In the first place, the eight characteristic elementary functions already determined are reevaluated in terms of rough numerical values. At first, the traffic volume levels a6 to a1 shown in Fig. 2 are determined (procedure P35).

Subsequently, based on these traffic volume levels, the traffic volume function C(t) of each set is converted into the traffic level function CV(t), to thereby prepare an ordinal array of eight elementary functions each comprising a plurality of sets for representing the traffic demand in one day (procedure P36).

In the case of the example illustrated in Fig. 2, it is assumed that the traffic volume level function CV(5) has seven values ($06 to $00). However, this number may further be decreased, for example, to four.

Further, on the basis of the determined frequency distribution of values of the floor congestion intensity function V(t), a curve IV, shown in Fig. 26 is depicted, in which the distribution levels pt to p4 are provided (procedure P37). Through this procedure, a rough floor congestion intensity function VV(t) is prepared to allow the types of combinations of the individual characteristic elements to be limited to a certain degree. Next, an array of the rough functions representing the corresponding elements of the traffic demand detected during a predetermined period is prepared for the purpose of extracting the typical characteristic modes (procedure P37).

Further, a characteristic mode extracting function PSm is determined for each of the rough characteristic modes (procedure P38), where m represents the number assigned to the characteristic modes which are recognized to be identical with one another.

By way of example, the above function PSm may be given by the following expression (23): PS(?=T(m){kl(CV(m?)+k2(W(m6 +vV()1 (23) The characteristic mode classifying affix m may be assigned to the characteristic modes in the sequence of detection. The function Tm represents the number of times the characteristic modes identical with or similar to the mode m are detected or alternatively the accumulated time.

If the frequency of same or similar characteristic modes exceeds a predetermined number of times, the characteristic mode which is detected earliest and whose function Tm is of the small value is included in the other characteristic mode of the closest proximity.

The recognition as to the identity of characteristic modes may be made on the basis of determination as to whether the associated congestion-concentrated floor functions 15 or 1D1 are coincident or not. Alternatively, when the value of the floor congestion intensity function VI(t? in concern is smaller than a predetermined one or below a predetermined level, the rough congestionconcentrated floor function IVn may be considered as $oo. In this case, the recognition of the mode identity may be validated, for example, when all of the functions IVS to IVg are in coincidence (procedure P37).

A plurality of the characteristic mode extracting functions PSn determined in the manner mentioned above are subsequently compared with one another, whereby the most significant one set or significant plural sets are selected from the newly recognized characteristic modes to be provisionally registered as new characteristic modes for the elevator control (procedure step P39). In this conjunction, when the preselected characteristic modes occur only rarely or does not occur at all, the number of the extracting or sampling times is increased so that the maximum number of samplings or extractions are effected to establish the characteristic modes for the elevator control as early as possible.The items of the characteristic modes to be provisionally registered are not the rough function values of the floor congestion intensity function W(t) and the traffic volume level function CV(t) but the values of the original functions V(t) and C(t). Then, discriminative identification performed later on can be made with an increased accuracy.

Through the procedures described above, the traffic characteristic modes present in the traffic demand inherently to the elevator system under consideration could be automatically extracted.

However, the present invention is never restricted to such particular mode extracting process. For example, several ten prospective characteristic modes may be previously prepared. Among them, those modes which occur in reality at a significantly high frequency may be selected as the characteristic modes to be utilized in the control of the elevator system in the manner described hereinafter.

Next, with reference to Fig. 6, description will be made on a characteristic mode preference function PP(m) for determining the preference or priority between the new characteristic mode automatically extracted and provisionally registered and the characteristic modes registered in precedence in conjunction with a characteristic mode creating control for creating characteristic modes by correcting or modifying the values of the registered characteristic elementary functions of the registered characteristic modes (procedure P40).

In this connection, it should be mentioned that the registered characteristic mode elementary functions include in addition to the detected characteristic mode elementary functions CVm, I and Vm of the traffic demand mentioned hereinbefore a periodical function TPm for learning elements or events which occur repeatedly with periodicity and a time function TH for learning those elements or events which take place periodically at predetermined time points each day.

In the first place, learning is performed for each of N1 sets of the characteristic mode elementary functions prepared through the procedure P33 as to which of spatial points Pm having polydimensional vectors and expressed by M1 sets of the registered characteristic mode elementary functions, the prepared characteristic mode elementary function lies in the closest proximity to.

Weighting of the individual elements is realized with the aid of constants k3, k4 and k5.

The vectors of a spatial point Pn formed by vectors defined by the detected characteristic mode elementary functions of the traffic demands are given by the following expression (25):

A spatial point Pm may be expressed similarly. A scalar quantity between the two points is determined in accordance with the expression (26) (procedure step 41). With regard to the width or duration of the congestion concentration or intensity, evaluation is made only in association with the first floor for simplification of the expression, since only the principle is of concern. In practice, however, it is preferred that evaluation should be made for the second and the third floor by using the correspondingly decreased weights. Further, difference in the floor number (difference value =lSnlmS) may be determined. In this case, the expression (26) is so altered that "0" is given upon coincidence of the floors, while "1" is given upon incoincidence between the floors.

In accordance with the expression (26), the number m of the registered characteristic mode having the closest spatial point Pm is determined and recorded as the registered characteristic mode of the closest proximity (procedure step P42). Namely, MiN(Pn,1,Pn,2, Pn,M,) (27) The procedure mentioned above is executed for n=1 to n=N1, the results of which are successively recorded in correspondence with the traffic demand characteristic elementary functions detected in succession (procedure P42).

Next, an accumulated value of time or number of times selected, for example, on the registered characteristic-mode (Pm) base is determined and used as an elevaluation function fTm for the registered characteristic mode (procedure P44). With this function, the registered characteristic mode is evaluated in terms of the frequency of its occurrence. It will be more preferable that the evaluation function be expressed in a form of the sum of products of magnitudes of traffic demand vectors.

Subsequently, a characteristic mode extraction preferance function 4iPm is determined for each of the registered characteristic modes in accordance with the following expression: Pm=( 1 k6)Pm+k6XTm (28) Among the characteristic modes, the one having the minimum value or those having less significant values are eliminated (procedure P45). Accordingly, the characteristic modes provisionally registered at earlier time are more likely to be eliminated because these modes are unlikely to assume large values.

Magnitudes of the individual elementary functional values of the registered characteristic mode functions Pm (M2 in number) finally determined are learned and established through the exponential smoothing in accordance with the expression (28) (procedure step P36).

As to the traffic volume function Cm or the traffic volume level function CVm, for example, the values of these functions for the traffic volume elements are created through learning (smoothing process over a long period) and calculation in the manner similar to the execution of the expression (28) on the basis of the elementary values of the characteristic modes approximating to the characteristic modes extracted during the predetermined period (one day or a week) or the elementary values determined through the weighted mean of the traffic demands decided as the characteristic modes m and the elementary values of the registered characteristic modes. As to the floors, the second and third floor functions are selected in the order of frequency of the occurrenced on the basis of data inclusive of those in the past.

Further, the time functions TPm and THm are also determined through the exponential smoothing in accordance with the expression (28) (procedure P47).

The period is learned by monitoring individually the time elapsed from the preceding detection to the succeeding detection to thereby recognize a plurality of the periods which occur at high frequencies, whereby the characteristic modes which are repeated in two different periods can be learned. Further, a number of time points in a day at which the event in concern take place at high frequencies are individually learned to be recorded as the time point function THm, with the result that more accurate predicted control can be performed.

On the basis of the prepared characteristic modes (hereinafter also referred to simply as the characteristics) which are established through extraction and learning in respect to the elementary functions thereof in the manner described above, the traffic information is recorded on the characteristic (mode) base through the procedure P50 shown in Fig. 4. Based on the information or data, optimal programs are selectively determined through the learning of the optimal control parameters for each characteristic mode and simulations performed by a computer incorporated in the elevator system if it has the capability of such simulation or by a computer installed for supervising the elevator system and others of a building or a large-scale computer installed at a control maintenance station, whereby the selected optimal programs are recorded (procedure step P60).

Subsequently, the present state of the traffic situation is analyzed in respect of the characteristic modes through the processes similar to the procedures P31 and P32, and the characteristic pattern recognition and learning are effected through the processes similar to the procedures P41 and P42.

Here, a supplementary description of a slightly varied and improved example will be added. In the case described in the foregoing, no consideration has been paid to the element of time. However, in the case of actual operation, delay is involved in the operation of the elevator control system and particularly in the group supervisory control for the elevator system between the exchange of the control algorithms and parameters and initiation of the control in accordance with the updated algorithm and parameters. In practice, this delay will amount to about 10 minutes in consideration of the fact that one cycle time of the elevator system is about 120 seconds on an average. In other words, the elevator control will generally be stabilized after the time elapse of 10 minutes from the initialization.

Accordingly, continuity of time should preferably be taken into consideration as the evaluation element in the characteristic recognition mode in accordance with the expression (26), (27) etc. For example, in connection with the expression (27), only the term of the scalar quantity Pt,m of the characteristic mode m, selected in the preceding cycle and being now utilized for the elevator control may be modified to (Pt,m)k6 where the factor k8 is of a value smaller than 1. Then, the characteristic mode selected in the preceding cycle and being now effective in elevator control will become correspondingly easier to be recognized. Second, the similar measures may be taken for the characteristic mode which is repeated at same time points every day, to thereby detect said characteristic mode as early as possible.

By way of example, it is assumed that a peak in the traffic volume which occurs about a quarter past eight on weekdays as in the office-going time zone shown in Fig. 1 is extracted as one mode Pm of the characteristics and learned, wherein the data of 08.1 5 is recorded as the time elementary function THm. Then, only the associated characteristic term of the expression (27), for example, may be replaced by the following expression, to thereby allow the characteristic mode to be identified earliest possible (procedure P70) and make the elevator operation be so adapted as to operate on the basis of the traffic data learned in the past days (procedures P50 and P60).

k7|tTHml Pt,m=(Pt,m) (1 kso k7 ollt-TH,l < k,l) (29) In this connection, assuming that k7 is, for example, 15 minutes, then the value of the characteristic term P,t,,m becomes smaller than the scalar quantity P(t).m within +15 minutes of the value "08.1 5" of the elementary function THm representative of the predicted time learned in the past days. Upon coincidence of the time, the characteristic mode in concern becomes more likely to be selected by a factor of (1k8).

When the characteristic mode identified through the characteristic mode discriminating process (procedure P70) described above approximates substantially to the characteristic Pm analyzed and learned (that is, when the values determined in accordance with the expressions (29), (26) and (27) are smaller than the predetermined values), the elevator operation is controlled on the basis of the control data determined by the characteristic Pm (i.e. the data prepared through the procedures P50 and P60).

When the values mentioned above exceed the predetermined values due to changes in the layout of a building or environment thereof, involving significant changes in the traffic demand, the first characteristic as well as a plurality of characteristics in the relatively close proximity thereto are discriminatively identified, whereby the control data prepared for each of these characteristics are used through interpolation or sum of the traffic data weighted in dependence on the proximity is determined (procedure 60) to thereby allow the elevator system to be controlled in accordance with the determined parameters (procedure P75).

However, when an event or festival is held on a particular day, the contents of that event as well as the time at which it takes place are previously inputted with the aid of switches indicating the associated floors and the types of control or with a keyboard or the like. Through analysis of the inputted data, it is determined which of the floors are to be serviced and what kind of control (e.g.

assignment of cars for service to the floor in concern with preference over other floors, extension of the door-open time, non-stop, permission of registration by using ciphers and others) is to be performed.

Upon entering the predetermined time zone for the event or festival, operation of at least a part of the elevator cars is carried out in accordance with the contents or data inputted previously in preference over the operations determined through the procedure P75 (procedure step P80). In this way, operation of the elevator system is controlled in accordance with the operation schedule (algorithm) by using control constants (parameters) which are expected to assure the optimum control for the elevator operation as demanded (procedure P90).

Further, the operation schedule for dealing with the prospected great changes in the traffic demand due to alteration in the layout of the associated building, holding of an exhibition for one month or the like may be inputted with the aid of a keyboard or the like, whereby the contents of the schedule may be provisionally stored in terms of the characteristic modes in the manner similar to that described hereinbefore in conjunction with the procedure P20, to thereby speed up the learning (procedure P95).

From the foregoing, the procedures for realizing the principle of the present invention illustrated in Fig. 4 will now be understood. In the following, some supplements will be made with regard to the matters in general.

(1) In general, after completion of the procedure P95, establishment of the characteristic modes through the routine P95-2 is regained. However, in the situation where the adaptive learning control is prevented from being carried out due to those factors which can not be included in the optimum control of the elevator such as the easy use, the routine P95-1 may be selected in place of the routine P95-2. In that case, the latter is taken only when command is issued by a schedule setting device.

Alternatively, the processing may usually proceed with the routine P95-2. In that case, however, it is preferred that the active operations in the execution of the procesures P40 and P60 should be restricted and that the devices used in connection with the procedures P80 or P90 should be implemented in an intelligent terminal so that the results of the learning can be displayed on CRT or the like in the form to be readily recognized by a person to thereby allow the elevator supervisor to confirm the contents as learned or modify a part thereof for re-registration. In this way, more active or positive control for learning can be realized.

(2) The procedures illustrated are for controlling the learning. In actual operation, collection and detection of the traffic demand data included in the procedure P30, identification or discrimination of the characteristic modes through the procedure P70 and the control of elevator operation through the procedures P75 and P90 are performed in parallel with one another.

In particular, the control of the elevator operation is generally required to be instantly activated at any time. Accordingly, the system must naturally be so arranged that the operation control can be activated in parallel with the other procedures. In the case of a computerized control, for example, those programs for the procedures which require a lot of processings such as the procedures P40, P60 and P75 may be allotted with tasks of lower level than that of the procedure P90, to thereby allow the elevator control program to be executed with priority or preference over the others, while the control for the learning may be carried out during the idle period.

In another modification, the learning control may be realized by making use of another computer and processed in parallel with the other procedures.

Next, effects resulted from realization of the principle of the present invention will be described by referring to Figs. 7 and 8.

In Fig. 7, a curve C3 represents changes in the traffic demand only for the visitors on a building such as a broadcasting center building which is abundant with the visitors in the daytime. The learning processes on the first, the second and the n-th day are illustrated on the assumption that no schedule is registered beforehand. Curves PLW, PLW' and PLW" represent magnitude of the time function THm for the visitors' characteristic modes. The value of PLW is zero on the first day and progressively increased at higher rate as is apparent from PLW' and PLW", resulting in that the time zone Pn during which the characteristics are discriminatively identified is selected at an earlier time correspondingly in dependence on the visitor's traffic volume.

Curve PLK, PLK' and PLK" represent the situations in which the characteristic elementary functions for extracting and establishing the characteristic modes for the visitors are detected.

Fig. 8 illustrates an example of inter-floor traffic or migrations of people inclusive of the visitors.

As will be seen from curves fUPUT,8 and fDNUT18 representing the alighting passengers' ratios at the individual floors, the number of the passengers leaving the up-travelling cars and the down-travelling cars at the fourth floor and those leaving the down-cars at the first floor are very remarkable.

Accordingly, corresponding characteristic elementary functions are learned.

Now, a first exemplary embodiment of the present invention will be described by referring to Figs.

9 to 11 showing circuit diagrams together with Figs. 1 2 and 1 3 illustrating data recorded.

An elevator operation control system (110) (adapted for performing the procedure P90) is composed of a hall call registration circuit 111, a device for defecting the number of passengers waiting for the cars at halls or landings (referred to simply as hall passenger number detector) 112, a car call registration circuit 117, a device for detecting the number of passengers (or weight) on the cars (referred to simply as car load or passenger number detector) 113, a door open/close control device 11 4, an elevator operation circuit 11 6 and an elevator driving circuit 11 5. These control circuit blocks 111 to 11 7 may be, respectively, realized in well known manners. (For example, a device for detecting the number of boarding and alighting passengers by using a car load or weight detecting device is disclosed in Japanese Patent Application No. 57027/1976.) A difference from the prior art resides in that the information available from these circuits, inclusive of the signals produced by opening/closing buttons and photoelectric devices, is made use of in a fine or delicate manner for the novel detection of the traffic demand according to the present invention. The other difference can be seen in the fact that means are incorporated for changing over the control algorithms and the control constants in response to the signals inputted to a traffic information detecting circuit 130 and finally fed back from an elevator operation control mode selector circuit 1 70 as the result of the learning (procedure P75).

A traffic demand signal Do 30 produced by a traffic information detecting circuit whose operation is illustrated in details in Fig. 5 is inputted to a traffic demand characterizing mode identifying circuit 1 50 (whose functions are illustrated in Fig. 6), while a characteristic-mode based information learning circuit 1 60 serves for accumulation or storage and learning of the data collected through cooperation of a traffic demand recording circuit D1 61An for recording data on the characteristic mode base (i.e. for each of the characteristic modes, separately), a service state recording circuit D1 61 Bn for recording service factors (such as the number of the elevator cars in operation, durations of hall calls, door-open time, noise level in building, erroneous riding, mischief, car boarding rejection, power source voltage, temperature and others) and a time zone recording circuit Dl 61 Cn for recording the results of learning such as detected time and periods (procedure P50). A clock time signal generator circuit 140 is further provided in view of controlling the operations of the other circuits.Additionally, there is provided in conjunction with the controls for the procedures P20 and P80 a reservation setting circuit 1 90 which is compared of a control mode recording circuit 192, a time recording circuit 1 91 and a control objective or item registration circuit 1 93 for recording the data as required in dependence on the contents inputted. Further, a desired value setting circuit 1 80 is provided which is composed of an energy saving degree command circuit 181, a service degree command circuit 182, an environmental factor command circuit 183 and an input device (interface circuit) 1 84 serving as the input means for these circuits 1 81, 1 82 and 1 83.The outputs of these circuits are supplied primarily to the control mode selector circuit 1 70 to be taken into consideration for the elevator control.

Referring to Figs. 1 0 and 11, extraction of the characteristics (modes) of the traffic demand which the present invention concerns in particular will be elucidated.

Through the input signal lines L1 11 to L1 17 leading out from the elevator control systems 110, data concerning the operation of the elevator and the traffic demand are monitored and accumulated (stored) in a data accumulating circuit D131.

The time point at which the data accumulation is started is recorded by a circuit Dl 31. The measurement of the traffic information mentioned above is effected by a circuit 131, whereby the traffic demand in the current state is arithmetically determined by a circuit 1 33 at every predetermined time interval (every several several minutes). More specifically, the traffic demand in concern can be determined by dividing a difference between the data of the circuit 1 31 recorded by the circuit D131T at the preceding time point and at the present time point, respectively, by the duration of the measurement.

The value in the present or current state is smoothed by an arithmetic circuit 1 34 imparted with a time constant of the order of several ten minutes. The output signal of the circuit 1 34 representative of the traffic demand in the current time zone is outputted on a signal line D134.

This signal is primarily utilized for stabilizing the characteristic recognizing operation (procedure P70) effected by a circuit 1 57 (Fig. 11).

The circuit 132 operates in response to an output signal Do 52 of a traffic demand elementary value calculation requesting circuit 152 shown in Fig. 11 which in turn operates on magnitude of the date D131 and the time duration or elapse, whereby the contents of the circuits D131 and D131T as well as the current time point are sampled and held by the circuit 132.

Subsequently, the circuit D131 is cleared of data, while the time point stored in the circuit D131T is updated for executing the collection of data for the detection of the succeeding traffic demand (inclusive of sampling of other information).

The signal D132 inclusive of the traffic demand detected in this way is supplied to the input of a traffic demand characteristic extracting unit (for processing the procedure P30) which is composed of circuits 151 to 156 shown in Fig. 11.

In the first place, the circuit 1 51 arithmetically determines the functions of the characteristic elements of the traffic demand on the basis of the signals Dl 52-2 outputted with some delay relative to the signal Do 52 every time the traffic demand is newly detected, the result of the arithmetic operation being recorded by the circuit 1 53 (corresponding to the procedures P32 and P33).

Subsequently, the circuit 1 54 performs the control operations for preparing the traffic volume level to form a function for evaluation (procedure P35) and determining a function for evaluating magnitude of the elements (procedure P36). This control is executed when a predetermined period, e.g. one day, has elapsed in case the current traffic volume (data signal D134) is low or otherwise at a time several hours before the lapse of the predetermined period. In this connection, the circuit 159 is provided with a view to accomplishing the control speedily by preparing previously the curve C2 shown in Fig. 2.

The recording circuit 1 53 records the data illustrated in Fig. 12 every time the traffic demand is detected. In the illustrated case, the n-th data D1 53, as recorded for one detection of the traffic demand is composed of 1 3 items which are sequentially recorded by the recording circuit configured in a chain-like form or a functionally loop-like form.

Here, supplementary description will be made about passenger's quality (character) discriminating elementary functions TM,n and TM2n which have not been described in conjunction with the principle of the invention.

Even in the same traffic demand, its nature will vary in dependence on the quality or character of the passengers and the time zones.

For example, the passengers tend to go about their business in a hurry in the morning while they will behave themselves so leisurely in the evening that the car reached first at the hall would be missed.

This kind of situation is taken into consideration in terms of the environmental function To12.

Further, the situation where children of light weight on an average occupy a large part of the passengers and the ratio of the mischievous calls, the ratio of calls by the passengers riding on wheeied chair and the ratio of VIP (very important person) calls are also taken into consideration in terms of the passenger quality (character) function TM2. Since these passenger quality functions represent factors of stress or burden imposed on the elevator control, they are additionally considered as the characteristic elements.

On the basis of the data recorded in the manner described above, extraction or sampling of the characteristics is performed (circuit 155), whereby the extracted (sampled) and learned characteristic modes are entered in a table of the recorded data, as is indicated by Do 53 in Fig. 12. The characteristic mode data in a number of M1 (1 8 in total in the illustrated example) listed in the table are classified into the characteristic mode record data Do 58 established by the schedule and the characteristic mode record data D156 collected through extraction and the learning (Fig. 11).

The schedule as well as data initially set are stored in C, RAM and/or ROM so that the characteristic modes can not be erased in the course of the automatic learning process.

In the foregoing, an exemplary embodiment of the present invention has been elucidated by referring to Fig. 9. However, the invention is not restricted to the illustrated embodiment but is susceptible to many modifications. For example, the portion composed of the circuits 151 to 1 56 and the circuit 1 59 for establishing and creating the characteristic modes shown in Fig. 11 may be provided independent of the elevator control apparatus.

By way of example, in the case of the system for controlling the elevator control circuit by means of a digital computer, the circuit portion for establishing and creating the characteristic modes mentioned above may be constituted by another digital computer or alternatively by a computer installed in an elevator supervising room.

Further, an transportable elevator maintenance apparatus such as disclosed in Japanese Patent Application No. 143513/1978 may be connected to the elevator control apparatus for a period as desired, whereby the changes in the elevator traffic demand are detected to establish correspondingly new or updated characteristic modes or creating the characteristic modes from the registered ones with the aid of the elevator maintenance apparatus which then serves as the apparatus for creating the characteristic modes for the elevator control.

In the following, a concrete embodiment of the present invention in which an operation control unit for an elevator system including a plurality of cars adapted to be driven in parallel and a control unit for dealing with the characteristic modes of the-traffic demand are constituted by separate computers, respectively, will be described by referring to Figs. 14 to 57.

In the description of this exemplary embodiment, a hardware arrangement for realizing the present invention will first be described, which is followed by the description of a software arrangement as a whole and the control concept contemplated by the invention. Finally, the software for realizing the control concept will be elucidated with the aid of table charts and flow charts.

Fig. 14 shows a hardware arrangement as a whole according to an exemplary embodiment of the present invention.

An apparatus MA for the group supervisory controls for elevator cars inciudes a microcomputer M1 destined for controlling the operations of the elevator cars operating in parallel and a microcomputer M2 destined for learning the traffic information for each of the characteristic modes of the traffic demand and performing simulation for each characteristic mode on the basis of the data obtained through the learning, as described hereinbefore. Between the microcomputers M1 amd M2, there is connected a serial communication processor or serial data adapter SDAc (described hereinafter) through communication line CMc for data transfer.

The microcomputer M1 for controlling the elevator operation is supplied with call signals from hall call devices HD through a parallel input/output circuit P 1 A and further connected to microcomputers E to En each installed in each of the elevator cars (it is assumed that the elevator system includes n cars) for controlling the opening and closing of the door, issuing the car acceleration or deceleration command and serving for other functions of the associated cars through serial communication processors (serial data adapters) SDA1 to SDAn and communication lines CM1 to CMn.

On the other hand, the microcomputer M2 is supplied through a parallel input/output circuit P 1 A with information signals required for determining through simulation the optimal operation control program as well as the relevant parameters and a signal PM produced by the input/output terminal PD which is used for establishing or registering the reservation described hereinbefore.

The car control microcomputers E, to En are supplied with the car call information required for the car control and connected through parallel input/output circuits P1A and signal lines SIO1 to SIO, to control input/output elements EIO1 to ElOn, respectively, each of which is composed of various limit switches for safety, relay, lamps for response and others.

A general feature of the present invention will be described with the aid of Fig. 14.

The elevator operation control microcomputer M1 incorporates therein an operation control program which is primarily destined for realizing the allotment of the calls. Upon execution of this operation control program, information required for the control is supplied from the car control microcomputers E1 to En and the hall calls HC. Further, among the information, those required for determining through simulation the optimal operation control program for each of the characteristic modes of the traffic demand are transmitted to the simulation-destined microcomputer M2 by way of the serial communication processor SDA. Further, in execution of the aforementioned operation control program, operation control parameters which are variable are made use of. More specifically, the operation control parameter includes, for example, a weighting factor indicating relationship between the waiting time involved in the evaluation function for the call allotment and the evaluation value of the power consumption, a time factor for determining the door open/close time and a parameter for selecting a control logic for the call allotment, i.e. control parameter for selecting the algorithm for the call allotment.

These operation control parameters can be arithmetically determined by the simulation-destined microcomputer M2 on the basis of the signal PM supplied from the terminal device PD and the data for simulation mentioned hereinbefore. Through this arithmetic operation, the optimal operation control program and the relevant parameter for the group supervisory control of the elevator system are obtained in dependence on the characteristics of the current traffic demand periodically or every time the new traffic information is collected.

It is assumed, by way of example, that the setting through the terminal PD is so made that the waiting time is at minimum. Then, the characteristic mode of the current traffic demand in the elevator system is identified with the traffic demand being predicted through arithmetic operation on the basis of the collected traffic information. From the data thus obtained, the algorithm for the call allotment which allows the waiting time to be minimum and the relevant operation control factors are determined through the simulation, to be subsequently recorded as the optimal operation control program and the parameters for the traffic demand at that time.In this way, the group supervisory control for the elevator according to the present invention can rapidly accommodate the traffic or environmental conditions of the building which vary from time to time, to thereby enhance significantly the group supervisory control performance for the elevator system.

Next, hardware structures of the individual microcomputers will be described in concrete. These microcomputers can be realised in simplified configuration as shown in Figs. 1 5 to 1 7. The MPU (micro-processing unit) which is the heart of the microcomputer may be of 8-bit or 1 6-bit capacity. In particular, the car control microcomputer (E,,. .. En) can be conveniently constituted by the MPU of 8bit capacity in view of the fact that the car control microcomputers need not be of a great processing capability. In contrast, the 1 6-bit MPU exhibiting excellent operation capability is preferred for the elevator operation control microcomputer M, and the simulation-destined microcomputer M2, since complicated arithmetic operations or calculations have to be executed by them.As the 8-bit MPU, there may be mentioned "HD 46800D" manufactured by Hitachi Manufacturing Co. and "28085" available from Intel Co. For the 16-bit MPU, "HD 68000" of Hitachi Manufacturing Co., "18086" manufactured by Intel Co., "Z-8000" manufactured by Zilog Co. or the like may be used.

As is seen in Figs. 1 5 to 1 7, connected to each of the microcomputers through a bus BUS are a ROM (read-only memory) for storing the control program, data of specification of the elevator and others, a RAM (random access memory) storing the control data, work (random access memory) storing the control data, work data and others, a parallel input/output circuit or PIA (peripheral interface adapter), and a processor or SDA (serial data adapter e.g. "HD 43370" of Hitachi Manufacturing Co.) destined only for the serial communication with the other microcomputers.

In each of the microcomputers M1, M2, E, to En, the associated RAM and ROM may be constituted by a plurality of elements in dependence on the size of the program and other factors.

Referring to Fig.16, hardware of the terminal PD may be realized in a structure similar to that of a test operation apparatus for elevator disclosed, for example, in Japanese Patent Application No.

82042/1978. Through a control console, the intended service, the desired power consumption level, the reservation etc., are inputted, while the contents of the input data as well as the relevant information are displayed on a CRT. The signal PM is fetched by the RAM through the data communication effected by way of the PIA and SDA mentioned above.

Upon next cycle or upon setting, cancelling or modification of the event, schedule, characteristic mode or the like, the data learned currently may be displayed on the CRT for the reference and additionally be printed out by a printer for preparing a report. To the latter end, a hard copy graphic printer associated with the CRT may be employed.

Referring to Fig.17, elevator control data such as the signals produced, for example, by a car-call button BC, a safety limit switch SWL, a relay contact SWRy and a car weight signal Weight are loaded in the RAM through the PIA. On the other hand, the data signals resulting from the arithmetic operation by the MPU is supplied to the control output elements such as the response lamp "Lamp", the relay "Ry" and others.

A hardware structure of the serial communication processor or serial data adapter SDA provided among the microcomputers shown in Fig. 1 5, 16 and 17 7 is illustrated in Fig. 18. As will be seen in this figure, the SDA includes as the main components a buffer for transmission TX8, a buffer for reception RX8, a P/S converter for performing parallel/serial conversion of data, a S/P converter for serial/parailel conversion of data and a controller CNT for controlling the timing among the components mentioned above. The aforementioned transmission buffer TX8 and reception buffer RXB can be freely accessed by the microcomputer for writing and reading of data.The SDA serves for the function to automatically transmit the contents of the transmission buffer TXB to the reception buffer RXB of the other SDA under the control of the controller CNT. In this way, the microcomputer is utterly free of the processings for the data transmission/reception and this can be used only for the other processings. By the way, the details of the structure and operation of the SDA are disclosed in Japanese Laid-Open Patent Applications No. 37972/1981 and No. 37973/1981.

Next, description will be made on the structure of software. In the first place, a general arrangement of the software will be elucidated by referring to Fig. 6.

As is shown in Fig. 19, the software may generally be classified into an operation software SF1 for performing the elevator operation control and a learning software SF2 for collecting the traffic information on the characteristic-mode base and performing the learning control through simulating means and the like. The former is processed by the microcomputer M, shown in Fig. 14, while the latter is processed by the microcomputer M2.

The operation control software SF1 is composed of an operation control program SF14 for directly commanding and controlling the elevator operation through the group supervisory control such as the call allotment processing, dispersion and stand-by processings of the cars and others. As the input information sources for this program, there are an elevator control data table 210 containing the data of the car positions, travelling direction, car call and others transmitted from the car control program (incorporated in the microcomputers E,--E, shown in Fig. 14), a hall call table 219, an elevator specification table 225 containing the number of the cars to be supervised etc., the optimal operation control program and the relevant parameters resulted from the execution of the learning software SF2.

On the other hand, the learning software SF2 is constituted by the programs mentioned below.

(1) Traffic information collecting program 230: The hall calls, the contents of the elevator control data table, data of the boarding and alighting passengers for each car detected at each floor and the like are sampled periodically on the on-line base for every predetermined traffic volume or with a predetermined period to thereby collect the data for arithmetically determining the program for learning the various traffic information through simulation and the relevant parameters. This program serves to collect mainly the traffic demand information for each of the floors.

(2) Characteristic mode identifying program 260: As mentioned hereinbefore in junction with the principle of the invention by referring to Fig. 6, the significant characteristics which make appearance at an increased frequency in the traffic demand in the elevator system is learned, to create and establish the corresponding characteristic mode for the elevator control on a long period base, e.g. over one week or longer.

Subsequently, the characteristic elements of the traffic demand contained in the traffic information collection table 231 prepared through the procedure mentioned in the above paragraph (1) are analyzed to identify the one which is most approximating the aforementioned characteristic mode.

In this way, the traffic information is collected on the characteristic base (i.e. for each of the characteristics).

(3) Program 260 for determining (leaning) traffic information on the characteristic-mode base for the elevator control: This is a program for arithmetically determining the elevator control data in consideration of both the contents of the table containing the data sampled on-line through execution of the traffic information collecting program and the contents of said table in the same time zone in the past.

(4) Program 271 for arithmetically determining various indexes through simulation: Data are read out from the simulation data table 262 containing data learned on the characteristic-mode base and the elevator control constant table 225 to perform simulations each for a predetermined number of algorithm parameters and control constant parameters, the results of simulations being written in a curve data table 272 in the form of curves. As the curve data table, there may be mentioned an average waiting-time curve table, a power consumption curve table, etc.

(5) Program 273 for arithmetically determining the operation control parameters: The contents of the curve data table 272 and a desired value table 280 established and loaded through the terminal device PD are read out to determine and record beforehand the parameters 274 optimized to the environmental conditions for each characteristic mode.

The optimal operation control commanding parameters 274 include additionally a part of the learned traffic information table 262 prepared through execution of the traffic information learning program 260 for each characteristic mode.

(6) Operation control parameter commanding program 275: Data of the current traffic information collecting table 231, the characteristic mode data obtained through execution of the characteristic mode identifying program 252, the data representing the generation of the characteristic mode in the past recorded as a part of the traffic information learning data table 262 and a part of the schedule data table 290 loaded through the terminal PD are read out to be used in executing a characteristic mode preduction program 277 for predicting the characteristic mode prevailing at present or occurring in the near future.

In executing a program 275 for preparing the control parameters, the characteristic-mode based parameters belonging to the predicted characteristic mode and the learned traffic information or data are selected out from the respective tables 274 and 262 to prepare a command parameter table 276.

The data thus prepared is transmitted to the elevator operation control system by way of the SDA (serial data adapter) to be utilized in place of the elevator control factor table mentioned above.

The learning software SF2 described in the foregoing may be considered as one of the intelligence control in view of the fact that the software serves to evaluate the results of actual operation and determine automatically the manner in which the elevator operation is to be controlled in dependence on the evaluation of the results.

In the foregoing, a general composition of the software according to an exemplary embodiment of the invention in which the microcomputers are used has been described. Next, elucidation will be made about methods for procedures for arithmetically determining the optimal operation control parameters through simulation for each of the characteristic modes of the traffic demand.

(1) Outline of hall call allotting program: In the present state of the art, there is adopted a hall call allotting method according to which the state of services to the individual hall calls (waiting time) is supervised to allot or assign the generated hall calls to cars in consideration of the overall call services. According to this method, the waiting time is used in the evaluation function for the call allotment. For example, there is proposed a method in which the longest waiting time of the car-assigned hall call generated at the floor lying in precedence to the floor at which the non-assigned hall call is generated is made use of as the value for evaluation (given by the expression 30 mentioned hereinafter). According to another proposed method, a squared sum of the waiting times of the precedingly assigned hall calls is utilized as the value for evaluation.

Further, there is known a method in which the waiting time of the currently generated hall call is utilized as the evaluation value. However, since no positional relationships among the cars are taken into consideration by these evaluation values, competitive services by the cars will be likely to take place in the normal traffic demand state, to degrade the performance.

(2) Control parameters for hall call allotting program: For preventing the competitive operation, concept of the stop-call evaluating function has been proposed (reference may be made to Japanese Laid-Open Patent Applications Nos. 47249/1977 and 126945/1977). More specifically, the stop-call evaluating function Tc is determined in consideration of the hall call HC,~1 assigned to a car under consideration which call has been generated at the floor adjacent to the one where a hall call HC, is being generated as well as the car calls CCj, CCi+2 of that car. In consideration of both the evaluating function Tc and the waiting-time evaluating value mentioned above, a new evaluation function 0 is prepared.To define this function sX by an expression, the waiting-time evaluation value is represented by T while the factor for determining the weight of the waiting-time evaluation value T and the stop-call evaluation value Tc is represented by jB. Additionally, with a view to allowing the elevator car already assigned with the hall call to successively respond to new calls independent of the floor at which the hall call HC; in concern is generated, a stand-by state evaluating value TR is introduced. When the weight of the evaluation value TR is represented by , the waiting-time evaluation value T is given by T",= MAX(TS'1, TS2k . .., (TSnk) (30) TS",=y1-TS'1+y2 DHi1-TW1 (31) where TS represents the time predicted for the car k to reach the i-th floor, TWi represents duration of the hall call at the i-th floor or a total sum of the waiting time of the passengers on that floor, and y1, Y2 represent the weighting factors of TS and TW, respectively. The evaluation function 6 is given by i,=T,Tk+Tii (32) Tk ,=S (33) Th=aR (34) where p represents a weighting factor on the order of e.g.O to 40 for the stop call (i.e. the call to be serviced) generated at the floor adjacent to the one where the hall call is issued, and S represents the probability of stoppage and takes a value of 1.0 if there is present a call to be serviced while taking an appropriate value (OSS < 1 ) for a predicted call. In the case illustrated in Fig. 20, the value of S is shown on the condition that the predicted call is neglected. Further, R represents the stand-by state or level of the car and takes an appropriate value (05us1 ) in the absence of the assigned hall call. The weighting factor a of R is, for example, of O to 40.

By using the evaluation function Tc given by the expression (32), the stop call adjacent to the generated hall call is taken into consideration to prevent effectively the competitive operation among the cars.

When a load-concentrated operation control parameter e shown in Fig. 20 is equal to 2, the stop call evaluation value Tc is determined by the following expression (35) in consideration of two floors preceding and succeeding to the floor i at which the call is generated. Namely, TC=zpS5 X 1.0+1 0x0+30x 1.0+1 Ox 1 .0+5x0=45 (sec) (35) Accordingly, assuming that the waiting time evaluation value T is of same magnitude among the individual cars, the car having the greatest value of Tc is considered to be most suited for being allotted with the generated hall call in concern.However, limitation is imposed to the value of Tc so that it does not exceed the doubled maximum value of the weighting factor , to thereby prevent the generated hall call from being left in the waiting state for a long time due to an extreme load concentration occurring accidentally. This is especially important in the case where the power consumption saving command on the order of 10% or 20% is inputted through the terminal PD in the time zone or characteristic mode in which the traffic volume is relatively small, and thus the elevator operation is performed with the simulation parameter e being set to the value of 6 or 7.

(3) Relationship between control parameters involved in hall call allotting program and power saving: By using the evaluation function TR defined by the expression (34), the elevator Operation can be constantly adapted to the currently prevailing situation with the number of the servicing cars being reduced while no particular cars being rested, to thereby accomplish the power saving operation control with the ratio of occurrence of the long waiting as well as the average waiting time being reduced to a possible minimum.

The stand-by level (state) may take the values mentioned below: O When the number of the allotted hall calls is greater than one, inclusive, R=1.

O When no hall call is assigned while the car call is present, R=0.1.

O When no service call is present, R=0.3 or greater.

O In the stand-by state, R=0.5 or greater.

O In the power saving state (stand-by state of an lengthened duration), R=1.

Accordingly, when no service call is issued as in the case of the car No. 1 shown in Fig. 20, the value of R may take one of 0.3 to 1, while R equals zero for the car No. 2.

Now, considering the waiting time T and the weighting factors a and jB of the stop call evaluation value Tc and the stand-by car evaluation value Tr in the expressions (32), (33) and (34), there exist values of a and p which are most effective for preventing the competitive operation and allowing the waiting time as a whole in the building (i.e. the averaged waiting time) to be reduced to a minimum.

On the other hand, as the factors a and p are increased, the car allotted with more stop calls is selected in preference over the others, resulting in that load is concentrated on the car in the servicing state, which means that the waiting time on an average is correspondingly increased. To say in another way, since the load on the other cars is decreased, the number of stops (the number of startings) of the cars as a whole is decreased to thereby reduce the power consumption.

(4) Introduction of simulation parameters for simulation: Examples of the relationships mentioned above are shown in Table 1 and Fig. 22 which are obtained through simulation on the conditions that the building under consideration has 1 3 floors, the elevator system has six cars, and that car speed is 1 50 m/min. Values of the weighting factors a and p are referred to as the load-concentrated operation control parameters. Simulation is performed for the case where simulation parameter e determined by the values of the factors a and is equal to 7.

Table 1

Simulation Average Power waiting time consumption Parameters (second) {KWH) e=1 &alpha;=5, =30, 5, 0,... 23 68 e=2 &alpha;=10, =30, 10, 5, 0 16 67 e=3 &alpha;=15, =3 5, 17.5, 12.5, 2.5 15 67 e=4 &alpha;=23, =40, 20, 15, 10, 5 25 63 e=5 &alpha;=34, =45, 22.5, 17.5,... 26 56 e=6 a=45, p=50,25, . . . 32 48 e=7 a=58, P=5 5, 27.5 As is shown in Fig. 22, an average waiting time curve fT and a power consumption curve fp can be obtained by varying the simulation parameter e for the load-concentrated Operation control. From these curves, it will be readily understood that there exists a minimum value of the average waiting time and that the power consumption is decreased as the parameter e is increased, while the average waiting time is correspondingly increased.

(5) Method of determining the optimal parameter: The above are the results of the simulation performed on the condition that there exist the traffic flows for respective destinations or destined floors (hereinafter simply referred to as the destined traffic flow or volume) from each landing. However, the destined traffic volume changes from time to time.

For example, the destined traffic volume in the office-going and office-leaving time zones utterly differs from the usual traffic mode. More specifically, although there exists an ordinary traffic volume both in the up (ascending) and down (descending) directions in the daytime, the traffic in the down direction occupies a major part of the overall traffic volume at the office leaving hour. Further, when there is a change in the tenants of a building in concern, the destined traffic flow pattern will be correspondingly varied. Accordingly, when the destined traffic volume A and the traffic demand B are simulated through the similar procedures as those described above, average waiting time curves and fTa shown in Fig. 21 are obtained.In this figure, the minimum values of the average waiting time lies at points Q and (i), respectively. With regard to the factor a, era=3.0 in the case of the curve fTA and web=2.0 for the curve fTB. It will thus be understood that the load-concentrated operation control parameter e should preferably be changed in dependence on magnitude of the destined traffic volume in order to reduce the average waiting time.

(6) Determination of parameter in dependence on the intended control objectives: Next, in conjunction with the power saving illustrated in Fig. 22, a method of arithmetically determining the parameter in concern will be described on the assumption that objectives for the control have been previously set. It should be understood that the similar control may be performed for the purpose of decreasing the ratio of the lengthened waiting time and optimizing the service completion.

It is now assumed that the average waiting time curve fT and the power consumption curve fp are given and that the desired value (objective) PM of the power saving is set at 1 0%. When the desired value of the power saving is 0%, elevator operation is performed with the load-concentrated operation control parameter e being set to e,(=3.0) which corresponds to the minimum point i) of the average waiting time curve. Accordingly, the power consumption is indicated by the point 0b . When the saving ratio of the power consumption at the point (6) is set to 10%, the reduced power consumption is indicated by the point Qc on the curve fp.Accordingly, the corresponding load-concentrated operation control parameter e is determined to be 4.5 (e2). Stating in another way, this means that the control for reducing the power consumption by 10% can be accomplished by setting the load-concentrated operation control parameter e at 4.5. In the exemplary case illustrated in Fig. 22, when the objective value of the power saving is set at a greater value, the average waiting time will be correspondingly increased. Thus, it is important to impose a limit on the objective value of the power saving so that the upper limit TLMT of the waiting time can be set at e.g. 30 seconds. For this reason, simulation with e < 6 is not executed since the waiting time of 30 seconds is exceeded when the parameter e is equal to or greater than 6.Thus, the evaluation index for the parameter e=7 is indefinite, as shown in the Table 1.

As will be appreciated from the foregoing, determination of various curves such as the average waiting time curve, the power saving curve and the like through simulation according to the present invention allows the value of the optimal load-concentrated operation control simulating parameter e to be determined in dependence on the desired or objective values as selected, whereby the loadconcentrated operation control factors a and p which constitute the elements of the parameter e can be readily determined.

(7) Learning of simulation parameter: By virtue of the introduction of the simulation parameter, there is obtained an advantage that the simulation can be accomplished at a significantly decreased number of times as compared with the simulation executed by varying individually a number of the elements which constitute the simulation parameter.

The exemplary parameter values listed in the Table 1 are stored in a specification table constituting a part of the work table for the simulation program 271. In this connection, it is necessary that at least the parameters a and p which are the constituents of the simulation parameter e=2 be contained in the elevator control specification 225, so that the elevator can be controlled on the basis of this value at the start of operation.

The basic specification is fed to the learning microcomputer M2 by way of the SDA (serial data adapter) to determine the values shown in the Table 1 and Fig. 22 through calculation. After completion of the simulation for learning the operation control parameter, simulation for learning the elementary parameters constituting this simulation parameter is performed.

(8) Introduction of parameters for selecting plural algorithms: In the foregoing, description has been made in connection with the method of creating the control parameters for the elevator operation control program having a predetermined algorithm. In the following, a method of creating parameters for both elevator operation control program having different algorithms and control constants used therefor will be elucidated.

From the foregoing description, it will be seen that the control constant parameters should be changed for each of the destined traffic flows through the learning adapted to the objectives in order to accomplish the object such as the reduction in the average waiting time.

Under the circumstance, the algorithm of the evaluation function for the call allotment must of course be taken into consideration. For example, the curve indicating the indexes for the control objective such as the average waiting time curve becomes different in dependence on the algorithm for evaluating the waiting time such as given by the expression (32). Accordingly, in order to accomplish the reduction in the average waiting time and the power consumption as well as the decreased ratio of occurrence of the lengthened waiting, the parameters of algorithm and the control constants must be selected to be most optimal for the destined traffic flows (which approximate to the floor-based traffic volumes of boarding and alighting passengers) as obtained through the learning of the traffic information collected on the characteristic-mode base or the time-zone base.

For example, the weighting factor P which determined magnitude of the stop call evaluation value represented by the expression (32) should preferably be selected at different values, respectively, in the up direction and the down direction, to more advantageous effect. This applies true in particular for the case where the ratio is 2:1 between the up direction and the down direction.

Further, when an elevator car allotted with a relatively great number of the floors to be serviced is releatively less crowded, it would be more favorable to the passengers to determine the value T of the waiting time evaluation function on the basis of the service-completed time. In particular, in the case of the algorithm for simulating the situation where a command for restricting the number of the passengers is inputted as the environmental factor command through the terminal PD and where the traffic volume is relatively small, a service-completed time curve should be prepared in addition to the curves fT and fp shown in Fig. 22 through the simulation using in place of the load-concentrated operation control parameter e shown in the Table 1 a simulation parameter y constituted by the parameter elements such as the weighting factors y, and y, of the below mentioned expression (36) which defines the service-completed time TAUT.

(9) Algorithm for evaluating waiting time on the service-completion time base: The service-completed time is given by ThsTk=3,3 . TH:+y4 TCkj (36) where the waiting time TH at hall and the car riding time TC may be arithmetically determined in the manners mentioned below, by way of example.

(l) Case 1. The service-completion time for the hall call at the landing of the i-th floor is determined in accordance with expressions (37) and (38) mentioned below: THkj=TSk (37) where TSk is determined in accordance with the expression (31).

TCk=J,5 TCS (38) where TCSj represents a predicted car-riding time calculated on the assumption that the car cali~is allowed from the i-th floor to a floor where the probability of stop in the same travelling direction as the hall call from the i-th floor is great, or to a terminal floor, and y, represents a factor of the probability of the car call being allowed to the floor preceding to the terminal one. This factor y, is a depending parameter to be determined as a result of simulation.

O2 Case 2. Calculation of the service-completion time for the car call to the i-th floor is performed in accordance with expressions (39) and (40) mentioned below: THk=3,6. TSkIH (39) where TS represents a value of the waiting time evaluation function TSkj determined in accordance with the expression (31) for the i,-th floor serviced immediately before generation of the car call to the i-th floor.

TCk=TSk+3,7. TCk (40) where TSkj represents a predicted time taken for the car No. k to reach the i-th floor in response to the car call to the i-th floor and entered in the same table asTSk appearing in the expression (31), and TCWk represents time durations taken by the cars for servicing the car call to the i-th floor as of the registration of the car call, and is stored in an especially prepared table for effecting the evaluation in terms of the service-completion time.The prediction time TICS, given by the expression (38) may be prepared by learning the values of TCWk averaged through statistic processing instead of the method described above The service-completion time TAST given by the expression (36) is determined for the registered car calls and the assigned hall calls which have been generated in precedence to hall call to be assigned.

Among the values thus determined, the maximum one is selected as the waiting-time evaluation value for the car in concern.

In other words, the evaluation value Tk given by the expression (32) can be rewritten as follows: TkI=MAX(TAsTki, TAsThk . . .r TASTY) (41) where n represents the number of the service calls located before the i-th floor and assigned to the car No. k.

(10) Calculation of algorithm determining parameters: For determining the evaluation function # by utilizing the service-completed time data, the expressions (32), (33) and (34) can be used as they are, with all the parameters y being set to 1 (one).

Calculation of the parameter e may be made through the procedure described before in the paragraph (7) by preparing the service-completion time curve in place of the average waiting time curve.

In general, in case much importance is put on the service-completion time, the average waiting time tends to be lengthened. Accordingly, the three kinds of curves described above are prepared through simulation with the ratios 3,3 and 3,4 of the expression (36) being used as the simulation parameters, to thereby determine the algorithm and the relevant factors or parameters optimized for the traffic demand in concern.

Namely, provided that T4 3,7 Y=- = --I Y3+Y4=1 3,3 3,6 3,6+3,7=1 and that 3,5=3,3 the operation performed in accordance with an intermediate function between two different algorithms can be simulated with the aid of the simulation parameter y being varied from 0 (evaluation in terms of the hall waiting time) to 1 (evaluation in terms of the service completion time).The results of this simulation are illustrated in Fig. 23a. Further, in case of a command for restricting the co-riding of the passengers from different floors is inputted through the externally provided terminal PD in the situation often encountered in a hotel or the like where all the users from the individual floors are treated as the very important persons (VIP), the simulation parameter3, mentioned above is varied from 0 (zero) to a value greater than 1 (one), e.g. to 6 in the simulation. At that time, the parameter web=2 determined already is used as the parameter e for the traffic demand B.

Subsequently, in place of the power consumption curve fp shown in Fig. 22, a co-riding probability curve f5 representing the probability of stopping at intermediate floors before reaching the destined floor is prepared, as is shown in Fig. 23b. On the basis of the co-riding probability curve f5 thus prepared, the simulation parameter YB which satisfies the commanded ratio, e.g. 60% inputted as the non-stop travelling ratio, is determined, to finally calculate the control factor parameters 3,3 to 3,7 on the basis of the calculated value Of y,.

In the foregoing, the method of determining the parameters to be utilized for performing the optimal Operation control has been described.

Next a table configuation for the Operation control microcomputer M, according to an embodiment of the invention in which the microcomputers are used will be described by referring to Fig. 24.

Broadly, there are provided an elevator control table block 210, a hall call table 219 and an elevator control factor table block 225. As to the individual tables of the blocks description will be made in conjunction with the operation control program 220, as occasion requires.

At first, the program for the operation controls will be described. In this conjunction, it is assumed that each of the programs mentioned below is divided into a plurality of tasks which are managed by a system program for assuring efficient controls, i.e. the so-called operating system (OS).

Accordingly, any program can be freely activated by a system timer or other programs.

Figs. 25 to 29 show, in flow charts, programs for the operation control microcomputer M1. In the following description, some emphasis is put on the important programs such as a program for preparing a predicted car arrival time table, a call allotment program and the like.

Fig. 25 shows in a flow chart a system program SF1A activated upon activation of the microcomputer M, and destined for the tasks of relatively lower ranks.

The program SO 1 A is started upon activation of the microcomputer M, or upon completion of initialization in the microcomputer M2, to transmit at first the elevator control specification tables 225 (Fig. 24) to the learning-destined microcomputer M2 through the SDAC (step 22A1). In this way, a car equivalent or simulating program used in execution of the simulation program 271 and having the same specification as that for the operation control can be put into operation.

Next, in order to re-start the operation in accordance with the learned car control specification upon recovery from the shutdown due to the service interruption or maintenance, it is decided whether the data such as the learned parameters have been transferred to the operation control microcomputer Mr from the learning microcomputer M2 by way of the SDAC (at step 22A2). When no learned data are available for initiation of the operation, the procedure proceeds to the step 22A5 where the operation parameter table 225B storing the data for setting/selecting various algorithms and control factors used in execution of the operation control program 220 is prepared on the basis of the data contained in the elevator control specification table 225 which is stored in ROM or nonvolatile RAM.

Upon restoration from the service interruption, procedure proceeds to the step 22A4 where appropriate values are placed in the second operation parameter table 225B on the basis of the data transmitted from the learning microcomputer M2 through the SDAc after restoration.

Next, at a step 22A6, with a view to sending the values contained in the elevator control data table 210 to the learning microcomputer by way of SDA0 for the purpose of collecting the traffic information described hereinbefore, data of 1 5 bytes and the number (NO) of 1 byte for discriminatively identifying the blocks of data to be transmitted are loaded in the transmission buffer TX8 of the SDAC.

The above step is repeated every cycle to send the required data to M2. Next, at the step 22A7, information of the hall calls HC is fetched through the PlAto be compared with the data of the hall calls registered in the hall call table 219, to thereby determine if there present the registration of a new hall call.

The task of the hall call allotting program is activated only when it has been decided that the new hall call is generated (step 22A8).

With the arrangement described above, the important data as learned can be recorded in both the operation (M,) system and the learning (M2) system to assure an enhanced reliability.

To this end, the learned data 22C obtained from the learning microcomputer or M2 system should be stored in a nonvolatile RAM such as CMOS RAM provided with a back-up battery for protecting the data from being lost due to the power service interruption. Further, a parity bit or sum data may be used for data verification, to determine the validity of the learned data by checking these additional bit or data.

Fig. 26 illustrates in a flow chart a procedure for arithmetically determining the predicted arrival time of elevator at a given floor for determining the data on the basis of which the waiting time evaluation value is calculated. This program SF2B is periodically activated by the OS every one second to determine the predicted time taken, respectively, for all the cars to reach given floors starting from their current locations in both directions for all the floors.

Referring to Fig. 26, the loop number of elevator cars is set at a step E20. Steps E20 and Eel 20 indicate that looped processings are executed for all the elevator cars. At a step E30, the floor is set at the car position. Next, at a step E40, initial values are placed in a work time register TWK. The initial values may represent time (seconds) taken for the car to start from the door open or closed state or time required for the elevator car to start from the rest state or the like. Next, the floor is shifted by 1 (one) in the car travelling direction (step E50), and decision is made as to whether the set floor and direction conform to the position of the car for which calculation is initiated (step E60).If the answer of the decision step E60 is affirmative "YES", this means that the predicted arrival time table is prepared for one car. Accordingly, the procedure jumps to a step El 20 where the similar processing is repeated for all of the remaining cars. On the other hand, when the decision step E60 results in the negative, i.e. "NO", the time Tr required for the car to travel for a single floor is added to the time register TWK (at a step E70). This time register TWK is then set to the arrival time table (step Ego).

Sub-subsequently, it is decided if the car call or assigned hall call is present, i.e. if there are calls to be serviced by the car under consideration. If affirmative, the car stops. Then, the single stop time TST is added with a predicted door time DTj stored in a predicted door time table 225B5 at a location corresponding to the i-th floor on the direction base, said predicted door time corresponding to the time required for the door to be opened and closed and the passengers to get off and on. With the result of this addition, the time register TWK is updated. Alternatively, a door time control parameter multiplied by a single standardized stop time Tst2 derived from the learning process may be added to the time register TWK (step El 00). Next, jump is made to a step E50 where the processing described above is repeated for all the floors and the directions in the same order as the car makes the service.On the other hand, when the decision step E90 results in "NO", the time placed in the time register TWK is updated with the stop probability PS, on the i-th floor stored in the stop probability table (on the direction base), and with the predicted door time and the single stop time tSt. Alternatively, the stop probability multiplied by the door time control parameter and the single standardized stop time Ts may be added to the content of the time register TWK (step El 10). Subsequently, jump is made to a step E50 where the above mentioned processing is repeated for all the floors and the directions.The one floor travelling time T, and the single stop time Tsf used at the steps E70, El 00 and El 10 are given as the optimum operation control parameters by the software of the learning microcomputer system, while the predicted door time DT, and the stop probability PSi are determined by measurement through simulation and the statistical processing. Determination of the predicted door time and the stop probability will be described hereinafter in conjunction with Fig. 46.

At the end of execution of the instant program, a hall call allotment program is activated for the reassignment of the calls waiting for a long time (step El 30).

The roll of the predicted door time table 225B5 is not restricted to the above but it can be used for the door open time which plays an important roll in determining actually the operation control of the individual cars. More specifically, an automatic door closing permission signal for each of the cars are produced by the operation control microcomputer. Alternatively, the values of this table 225B5 may be sent to the microcomputers En provided, respectively, for the individual cars by way of the SDAn, to thereby determine the door open time limit of the cars. (To this end, a predicted door time included in the specification of an automatic door closing count timer may be used.) Fig. 27 shows in a flow chart a full car (full load or saturation) prediction program. At first, activation of this program will be described.The activation of the full car prediction program is effected through the same task as the hall call allotment program SF1 D described below. More specifically, the hall call allotment program is executed immediately after the activation of the full car prediction program so that error due to difference in time be reduced.

Referring to the flow chart illustrating the full car prediction program, initialization is first effected for the car K at a step G10 where K is set equal to 1 in this case. Subsequently, the number of passengers in the car (hereinafter referred to as the intra-car passenger number) is set to FK (step G20), which is followed by a step G30 where the hall call generating floor is set to a number i. It is determined whether there is present a hall call on the floor i (step G40). If no hall call is found, it is then determined whether there is a car call to the floor i (step G50).If the car call is present, a car-calldependent full car predicting value Pc, for the floor i (prepared for each direction) in a full car prediction table 225B3 is subtracted from the intra-car passenger number (step G60), which is followed by updating the floor number i by one in the car travelling direction at a step 6100. If no car call to the floor is found at the step G50, jump is made to a step G 100. When a hall call on the floor i is found at the step G40, it is checked whether there is a car call to the floor i (step G70).If the car call is present, a hall-call-dependent full car predicting value Ph, for the floor i (prepared for each direction) in the full car prediction table 225B2 is added to the intra-car passenger number FK, from which the aforementioned car-call-dependent full car predicting value Pci is subtracted (step G80), and jump is made to the step G100. When no car call is found on the floor i at the step G70, the hall-call-dependent full car predicting value Ph, is added (step G90), which is followed by the step G100 where the floor i is incremented by 1. Next, determination is made as to whether or not the floor i is the top floor or the bottom floor (step G). If the answer is "NO", jump is made to the step G40.On the other hand, if the answer is "YES" it is checked whether the value of the intra-car passenger number FK is within a fullcar load parameter learned by the learning microcomputer system (step G120). If the answer is "NO", the elevator car K is decided to be incapable of service (step G130). Thereafter, it is checked if all the cars have been examined (step G 1 40). If the answer resulted from the step G 120 is "YES", the aforementioned step G140 is repeated. When the above procedure has been completed for all the cars, this program comes to an end. Otherwise, a next car is selected (i.e. K is incremented by 1) at a step G150 and the step G20 is regained to repeat the processing described above.By the way, the car-calldependent full car prediction Ph, and the hall-call-dependent full car prediction Pc, employed at the steps G60, G80 and G90 are determined by measurement through simulation and statistical processing. Preparation of these full car prediction data will be described hereinafter in conjunction with Fig. 47.

Fig. 28 shows in a flow chart a call allotment program which is activated by a hall-call asigning task. In this program, it is assumed that the algorithm for the call allotment is, by way of example, so prepared that the long waiting call (i.e. the call waiting for the car service for a long time) is minimized, as indicated at a step H50. (This algorithm will be elucidated later on in conjunction with Fig. 29.) Through the steps H20 to H80 and H30 to H70, the looped processings are performed, respectively. At a step H40, it is decided whether hall call, if any, is issued. If no hall call is present, a step H70 is executed where the presence of the hall call is checked for all the floors and the directions. If the step H40 results in "YES", the long-waiting-call minimizing call allotment algorithm is employed to assign the optimal car to the call (step H60).

Fig. 29 is a flow chart illustrating the processing in accordance with the long-waiting-call minimizing call allotment algorithm. In order to decide which of the cars is optimum for servicing the call, looped processing is executed for the cars K through steps H50-1 to H50-7. In the looped processing, a predicted maximum waiting time Tmax for the assigned hall calls issued from the floors located in the car travelling direction, inclusive of the hall call being generated, is first determined at the step H50-2. In this connection, it should be understood that the predicted waiting time is a sum of a hall call holding time representing the time lapse from the generation of a hall call to the instant time point and the predicted arrival time.At the next step H50-3, the stop call evaluation value Tc is determined on the basis of the stop calls from the predetermined preceding and succeeding floors, inclusive of the hall calls being issued, which is followed by a step H50-4 where the stand-by state evaluating function T is determined to validate the stop call evaluation function 0 (where 0=TmaXTC+TR) through cooperation with the predicted maximum waiting time Tmax mentioned above (step H50-5). Then, the elevator car for which the value of the evaluation function 6 is minimum is selected (step H50--5). The processing described above is executed for all the cars K which are capable of servicing.Thus, the elevator car K to which the optimal evaluation value is applied is selected through execution of the step H50-6.

In the foregoing, description has been made on the main programs for the elevator operation control including the processing flows of the program for preparing the arrival time prediction table, the call allotment program and others. Additionally, as the operation control programs, there can be mentioned a plural car dispatching program for allowing a number of the cars to do service for the call of the congested floor, a dispersed stand-by processing program for dispersing the cars among respective predetermined floors in the stand-by state when the traffic demand is small. Description of these programs is omitted herein.

Next, configuration or arrangement of the table utilized by the learning microcomputer system M2 constituting another main part of the illustrated embodiment will be described in conjunction with the block diagram shown in Fig. 1 9.

(1) Traffic information collecting table: The traffic information table 231 as used is of a structure similar to the one shown in Fig. 12 and used in the first exemplary embodiment described hereinbefore. This table is used primarily for establishing and creating the characteristic modes themselves. Further, to prepare a traffic information storing table 256 mentioned below, a table similar to that 256G shown in Fig. 30 is made use of for detecting and sampling the traffic elements included in the characteristic modes.

(2) Traffic information storing table: Fig. 30 shows a basic structure of the table for storing and learning the traffic information. This table structure is an assembly of 1 8 table blocks which are classified in correspondence with the characteristic mode registering tables D158 and D1 56 shown in Fig. 13.

In practice, it is very difficult to impart the inherently proper names to the characteristic modes.

Only for the reference, nicknames of the characteristic modes which are prospected in view of the results of characteristic creating and learning process performed in an elevator system installed in a building occupied by only one company are entered with brackets. The concept of the characteristic modes is categorically utterly novel and differs from the hitherto known concept of the operation pattern in any respect. The characteristic mode by itself plays no important roll in determining the operation manner and specification (inclusive of the determination of the operation patterns), but serves for meaningful function in classifying the traffic demands which are the objectives for the learning and determination of the parameters involved in the elevator operation control.

There are provided a table 256G and others for accumulating the traffic information for each of the characteristic modes (i.e. on the characteristic-mode base). Further, there are provided tables for dividing some of the information into a plurality of data or for each of the floors (on the direction base) as is the case of a passenger number table 256G7 for the characteristic mode NO7 (M7).

Fig. 31 illustrates in detail a structure of the time zone record table 256G2 shown in Fig. 30, in which signals representative of the time points in concern in a day are recorded among the output signals of the CLOCK LSI 299.

In connection with the recording of the time points, it should be mentioned that time points at which discriminative identification of the characteristic mode (e.g. identification or recognition of the nickname of the characteristic mode M7, ordinary congestion) for collecting the traffic information is initiated and ended are recorded in a pair.

3) Schedule reserving table 290: Referring to Fig. 32, there is illustrated a typical structure of the schedule reserving table together with some examples of the recorded data. Content of an event held only once is externally inputted through the terminal device PD and recorded in an event reserving table 291. For example, in an assumed case where it is scheduled that several ten important visitors are prospected who use the car between the fourth floor at which the reception room is provided and the first or lobby floor, the time in concern as well as the floor numbers and the control mode VIP are designated (inputted) and recorded in the table 291. When the reserved time has elapsed, the recorded data is invalidated or erased. These recorded data are preferred to upon execution of a characteristic-occurrence prediction program.More particularly, the reserved time is compared with the clock signal produced by the LSI clock 299. Within the reserved period as determined through the comparison, the service command for putting the preference to the service for the fourth and first floors is issued.

In more concrete, the command parameter command that down call from the first floor both imparted with a preference level are to be served earlier by one minute or so in preference over the other floors is set in a command table 276. Further, with a view to lengthening the door open time so that the important visitors can get on the car arrived for service without fail, a command parameter is set at the table 276 which causes the data DT1 of the door time prediction table 225B5 for the down car on the fourth floor and the up car on the first floor to be set at a value which holds the door opened for six seconds, by way of example.

All the other schedule data concern the learning process and will be described, as occasion requires.

(4) Traffic information table included in information learning table 262: Fig. 33 shows structures of tables destined for recording information learned by the learning microcomputer M2 over a long time span. More specifically, there are illustrated an information table 262Z common to a plurality of characteristics together with tables 262A-1 and 262A-2 for the characteristic mode M1 as typical ones of the traffic information tables for the information each learned on the characteristic-mode base.

A floor-based average door open time table 262A-1 stores therein the learned values representing the averaged door open times determined on the basis of the door position signal fetched from the elevator state table 210A through the SDA on the floor base (for each direction). The contents of this table are utilized as the basic data for determining the stay or dwell time through simulation and preparing the door time commanding parameter described hereinbefore.

A car-based door open/close time table 262Z3 records therein the times required for opening/closing the door independent of the traffic demand on the car base (i.e. for each of the cars).

These times are measured through the similar procedure as mentioned above and utilized in the similar manner. The time under consideration varies in dependence of the width of door, regulated speed of the door driving unit and frictional resistance presented by the car door and the landing door. Further, the door opening/closing time may sometimes vary under the frictional resistance exerted by dusts cumulated around the door.

A floor-based special call (passenger) generation ratio table 262A1 3 records therein the learned ratio of the particular passengers who use the car with the VlP-only button call or cipher call or the wheeled-chair passenger call and/or passengers who use the cars through the call generated by especially provided button or cipher and serviced only by the cars which are capable of servicing the underground floor in the case where a half of the cars in a group are capable of providing the service to the underground floor and/or passengers who use the car through the destination floor register installed on the landing in addition to the ordinal buttons to permit a so-called port transportation.

The value recorded in this table 262A1 3 are utilized for validating the passenger quality discriminating function l(TM,n) of the table 1 zone shown in Fig. 1 2.

Further, a table 262A1 7 destined for recording the learned body-weight distribution and a table 262A1 7 for recording the learned distribution of the passengers on the typical floor (which is greatest in the number of the passengers throughout a day) on the arrival-time interval base are utilized for validating the other passenger quality discriminating function II (TM2n).

The two learnings for preparing the above tables can be realized in a facilitated manner by making use of the hall passenger number detecting apparatus 112 of the first embodiment shown in Fig. 9.

The three tables mentioned just above are not only used for the learning of the characteristic modes but contribute to a significant improvement of the learning performance according to the present invention.

(5) Table for recording learned characterstic modes and day-based patterns: A traffic demand characterizing mode table 262Z6 is realized in a structure similar to that of the characteristic mode registration table used in the first embodiment and shown in Fig. 1 3. Upon establishment or erasing of a characteristic modes, the elementary values thereof as well as the relevant time are recorded in a table 262Z7 having a similar structure as that of the table Do 59.

Further, even the registered characteristic mode is updated and recorded again, when the elementary values thereof change significantly as a whole. For identifying these characteristic modes, the characteristic mode numbers M1 to Ml 9 mentioned hereinbefore and additionally the characteristic mode supervising numbers P 1 to P255 for the recording over a long period (e.g. one or two years) are used. These numbers may be same as those attached to the associated recording tables. When the mode number attached to the characteristic modes in the ordinal sequence in which the modes are recorded (or occur) exceeds the number P255, the control is so made that the corresponding characteristic mode may be recorded at a recording area which is allocated to the oldest one of the characteristic modes not used at present. Prediction of these characteristic modes can further be improved by learning the characteristic modes on the day base. To this end, the day-based characteristic pattern is established through the similar process as the extraction and establishment of the characteristic modes mentioned hereinbefore and recorded in a table 262Z8 which has a composition shown in Fig. 34 where five typical day-based patterns are entered. It is expected that cancellation and addition of the day-based pattern is unnecessary in most cases. If otherwise, the daybased pattern would have to be rearrayed in the order from the day on which the traffic volume is smallest to the day where the volume is greatest.

It should be noted that the individual elements of the day-based characteristic pattern are subjected to the exponential smoothing in accordance with an expression similar to the one (28) with a time constant of several days.

Identification of the day-based patterns thus determined are effected by executing the program 257, the results of which are recorded in a table 262Z9 in the order of calendar days over one or two years. A structure of the day-based characteristic pattern recording table 262Z9 in which numerical examples are entered is illustrated in Fig. 35.

It will be seen that the data of January 4 in the next year is entered at the area allocated to the data of January 1 of the last or preceding year.

(6) Simulation result storing table 272: Fig. 36a shows an overall table structure for recording the results of simulation of the characteristic-mode base.

Tables 272A to 272J for the respective characteristic modes M1 to M9 correspond to the characteristic mode registration table Do 58 shown in Fig. 13, while tables 272K to 272T for the respective characteristic modes Ml 1 to M19 correspond to the characteristlc mode registration table F156.

In the case of the illustrated embodiment where the microcomputers are employed, these characteristic modes are registered in the table 262Z7.

Fig. 36b shows a concrete example of the table structure for recording the final output results of simulation of the characteristic mode M1.

These tables are updated every day. However, the characteristic mode making appearance only once in a week or those which occur in particular seasons such as spring and fall are stored in a special table prepared to this end so that the data of these characteristic modes can be utilized as occasion requires.

(7) Command parameter table 276: Fig. 37a shows a structure of a whole table for recording the command parameters supplied to the operation control microcomputer M1. Since most of these parameters have been mentioned herein before, the following description concerns only the remaining ones.

Details of load control parameters 276C are shown in Fig. 37b. The values of these control parameters are estimated for each of the prospected characteristic modes of the traffic demand by executing a control parameter preparing program 275 and recorded on the characteristic-mode base.

More particularly, these parameters are determined on the basis of the learned traffic data contained in tables corresponding to the table 262A9 which records a frequency distribution of the boarding rejection ratios on the car-load base, the table 262A8 which records a frequency distribution of the left-off passenger ratios and the table 262A1 7 for recording the distribution of body-weights of the passengers, all shown in Fig. 33.Further, desired power saving load tables 276B5 and 276B6 as well as allotment limiting load tables 276B3 and 276B4 are prepared on the basis of optimal values calculated from the data contained in the simulation parameter table 262A22 containing input parameters for determining the control specification of the simulation program 271, the power consumption curve table 272A3 for the characteristic mode in concern shown in Fig. 36 prepared on the basis of the results of simulation and average waiting-time curve table 272A1.

A power saving control in which load parameters are utilized is disclosed in Japanese Patent Application No.31195/1981.

In the following Table 2, there is shown examples of learned control parameters which have direct influence on the car load values CWk (Fig. 24).

Table 2 Learned values t Initial values 102% F 110% 75% v 80% 50% v 80% 70% F 80% 43% v 45% 47% r 45% 34% v 40% In the above table, the initial values are set equal to those contained in the elevator control specification table 225. For example, the full-car load value WBm may be set at a point where the boarding rejection ratio becomes greater than 1.5 times of the rejection ratio at the time of the light load. When the intra-car load exceeds the value WBm, the associated car control microcomputer En decides the full-car state, to thereby inhibit the response to the hall call.When the value WBm is selected excessively small, the transporting capability is lowered, while the boarding rejection ratio is increased when that value is selected excessively large. For example, it is conceivable that in case the traffic is relatively light, office girls, boys and those passengers who carry luggages will refrain from riding the arrive car in consideration of trouble to other passengers and will register again the hall call for calling the car which is expected ready to come to make service. In this conjunction, the phenomenon of the passengers being left off will occur frequently when the load value WAm of the arrived car is exceeded due to an excessively large number of the passengers.Thus, the load control parameters WBm, WCUm and WCDm have to be selected on the basis of the learned data and the results of simulation described above so that the undesirable phenomena can be evaded.

The parameters WAn and WB n are also transmitted to the car control microcomputer En by way of the operation control microcomputer M,. Further, the parameters WCUm and WCDm are made use of at the step G120 shown in Fig. 27. The overload parameter WAn may serve as a pointer for evading the overload condition. The load value at which the car is prevented from starting is fixed at 1 10% of the full-car load.

The stand-by control parameters serve for the functions corresponding to those of the conventional inter-floor dispersed stand-by control and the Operation in which all the cars are located in the stand-by state on the starting base floor. These stand-by control parameters command the number of cars required to be placed in the stand-by state on the floor base, the directions, the standby states of car such as door-opened or closed state, the door open time and the like.

Among the stand-by parameter, the parameter Nile) which commands the number of the cars required to be set in the stand-by state is determined in accordance with the following expression (41 A) on the basis of information about the floor-based frequency distribution of boarding passengers Phi,m,, the power saving requirement and the number of the currently operatable cars N above all among the traffic information learned on the characteristic-mode base.

where SPh mX represents the number of passengers on all the floors, and ,7 represents a weighting parameter which corresponds to a point where product of the curves fT and fp shown in Fig. 22 (aS,7 being taken along the abscissa) obtained through simulation from the values of the required power saving and the average waiting time is minimum. The parameter ,7 may also be determined in accordance with the following expression: a17= Ph"m/[unit time (5 min)/average round time]xO.5xprescribed number of car passengers (41 B) where the term of 0.5 which means the riding passenger ratio of 50% is effective in preventing the fullcar condition.This ratio should preferably be selected slightly smaller when a!,7 > 1 and set at 2 or greater in the large.

Example of the control for the stand-by cars ready for servicing the congested floors in the manner described above is illustrated in Fig. 61 for the typical characteristic modes of traffic demand, respectively.

In the case of the dispersed operation also illustrated in Fig. 61, when unbalance in the service distribution exceeds a predetermined level, control is so performed that the cars located in the serviceexcessive floor zone are moved to the service-scanty floor zone.

In this connection, the predetermined value serving as the reference to the unbalanced service condition mentioned above should most preferably be so determined through simulation that the product of the power consumption curve and the average stand-by time curve be at minimum.

The parameters commanding the floors to be serviced correspond to the control parameters which have heretofore been employed in the express services on the floor division base and the block express service or in designating the thinned-out floors and the junction floors and are utilized for commanding the possibility or impossibility of service to the car calls, designation of the cars capable or incapable of the service and the like on the floor base (for each direction). Since these parameters are of utmost concern for the people who use the elevator system, the changing of the parameters should be authorized by the supervisor. To this end, arrangement may be so made that the parameter changing request is inputted through the terminal PD, in response to which the supervisor issues authorization, whereupon the parameters are automatically changed.

The floor-based (direction-based) concentrated service parameters command the control of the allotment of plural cars, continuously attending service, changing-over of the service preference levels, extension of the door time, voice information such as "three more may get on" and the like. Novel features can be seen in the provision of the parameter commanding the setting of the congestion level for a floor to which a plurality of cars are dispatched in response to the hall call issued at that floor and one or more cars are constantly allotted regardless of the presence or absence of the hall call in dependence on the congestion level. When the congestion level is set high, either one of the two concentrated services mentioned above will not be performed.On the other hand, when the congestion level is set low for the floor on which congestion is expected due to the reservation for the event or learning of the past data which takes a lot of time, the concentrated service control will be initiated in response to a mere symptom of congestion or the issue of the concentrated-service commanding parameter.

The environment control parameter serves to prevent noise from being produced due to the parallel running of the cars and limit the top speed of car for preventing the noise generation.

Re-assignment control parameter determines the value of command for deciding whether the hall call once allotted is to be re-allotted or addition allotment be made. Further, this parameter determines the value of a command for changing over a service reservation indicating lantern, if present, to the car expected to be able to perform the service in concern.

Information control parameter serves to issue a command for inhibiting the service guide of same floor and direction, a command for changing over the direction of the service guide, and a command for limiting or inhibiting the audio or voice guide or limiting the volume.

An input control parameter is employed to command which one of the car carrying load detector and the intra-car passenger number detector is to be used as the sensor of the operation control system or alternatively both of the detectors are to be used. This parameter in turn is determined on the basis of the result of simulation and the value contained in the traffic information storage table in which the actual elevator operation data are collected.

In particular, it is determined which one of the above detectors can collect the traffic information providing the corresponding results of simulation approximating closer to the average waiting time and the long waiting ratio (or a by-5% longer waiting time), whereby the detector capable of providing the information closer to the simulated value is selected as the sensor. When both detectors are equivalent to each other in this respect, they are both used as the sensor.

Further, as regulating command is issued to sensors such as the hall passenger number detector and an ultrasonic door sensor whose sensitivities are required to be regulated.

As other parameters, there can be mentioned a time limit value parameter commanding the turnoff time of an intra-car illumination lamp, a turn-on control command parameter for a power saving ratio display lamp and others.

In the foregoing, structures of the main tables used in the learning microcomputer system M2 have been described. Although there are provided the work tables WK used upon execution of the associated programs and the control table used in the OS. Description of these table will however be unnecessary. Since the information learning table 262 is held in the activated state for the learning over a long time span, data holding means is required to prevent the data from being loss due to the power service interruption.

Next, programs of the learning software SF2 will be described by referring to Figs. 38 to 51.

This software is not divided into the functional blocks in the same manner as the software described hereinbefore in conjunction with Fig. 19 which shows the general structure thereof. This is for the purpose of simplifying the program and speeding up the processing. However, the invention is not restricted to the software structure described below.

Fig. 38 shows in a flow chart a system program SF2A which embodies mainly a traffic information collecting program 230 and a characteristic mode identifying program 252 and is adapted to be activated upon starting of the learning microcomputer M2.

At first, at the step 22A1 shown in Fig. 25, the content of the elevator control specification table 225 provided in the operation control microcomputer Mr and the content of the specification table incorporated in the car control program are read out through the SDAC to be placed as the initial value in the traffic information learning table 262Z shown in Fig. 23.

A car-based service floor table 262Z1 0 and a measured inter-floor running time table 262Z1 are prepared on the basis of the values read out from a floor height table and a rated speed specification table provided in each of the car control microcomputers E1 to En. An event reservation recording table is simply cleared. As described hereinbefore in conjunction with the step 22A2 shown in Fig. 25, the data obtained through the learning process are held as they are.

At a step 2B, the traffic information collecting program 28 is activated. Details of this step are illustrated in the flow chart of Fig. 39. Outline of the procedure is substantially same as the procedure P3 1 illustrated in Fig. 6.

Although illustration of the data required for this processing is omitted, it should be mentioned that a reception processing program corresponding to the step 22A6 shown in Fig. 25 is provided in the task activated periodically by the OS to receive the data such as the intra-car load data CWk among the data for the operation M, system, to thereby prepare a table equivalent to the one shown in Fig. 24 or at least to a part thereof.

In connection with this data transmission and reception, the DAM should preferably be used in place of the data transfer hardware SDAC to thereby realize the high-speed data transfer, since no processing will be then required in connection with individual application programs.

Further, shown arrangement is made such that the data used in both of theM1 and M2 systems are stored in a common memory by way of a common bus as is shown in Figs. 24 and 37, it is possible to read out data utterly freely from both microcomputers.

When a predetermined number of the passengers is exceeded or when a predetermined period has elapsed, the procedure leaves the step 2A and proceeds to a step 2C where the traffic information are collected on the characteristic mode basis (i.e. for each of the characteristic modes). Details of this step 2C are illustrated in the flow chart of Fig. 40. Outline of the procedure of this step is same as the procedure P42 shown in Fig. 6.

Collection of the traffic information through the steps 2B and 2C is performed for a predetermined period e.g. one day (step 2D). Subsequently, the learning control is effected through steps 2E to 2H.

The predetermined period mentioned above is not restricted to one day or 24 hours but may be one week. At the step, the creation processing of the characteristic modes themselves through correction, new establishment, deletion or the like (same processing as the procedures P35 to P39 shown in Fig. 5 and the procedures P41 to P47 shown in Fig. 6). However, the step 2E may be executed at a slower speed than the steps 2F to 2H destined for creating the control parameters. In this case, the traffic demand information used at the step 2E may be continuously corrected for one week, for example, and processed in a lot on a week basis, to thereby improve the performance as a whole.In a building having numerous and various characteristic modes in the traffic demand such as a hotel, a rental office building and an assembly-hall building equipped with various facilitates, the step 2D may be repeatedly executed for a shorter period than a day to thereby allow the sampling or extraction and selection of the characteristic to be realized more scrupulously.

In the thin time zone, deviation of the collection period from 24 hours would involve no problem.

At the step 2H, the tank is activated which includes the program SF2B for learning the parameters which issue commands to the operation control microcomputer system for each characteristic mode. In the following, the steps 2B to 2H will be described in detail by order with reference to the drawings.

In the traffic information collecting program 2B an example of which is illustrated in Fig. 39, the variables are first initialized with the collection table being cleared (step 2B 1). Next, same information as the one contained in the table 2566 shown in Fig. 30 is detected and placed in a traffic information collecting table 231 shown in Fig. 19 (and having a substantially same structure as the table 262A shown in Fig. 33), whereby the collected traffic information about the number of the passengers and others is stored (step 2B2).

The number of the boarding and alighting passengers may be detected in terms of changes in the intra-car load in the manner disclosed, for example, in Japanese Laid-Open Patent Application No.

1401 47/1 977. To this end, use is made of values CWk (hereinafter, k will be used as ordinal or array variables of the car microcomputers E1 to En) of the intra-car load (current intra-car load values for each of the cars) placed in the car control table 210. When the intra-car load at the starting from the last floor is represented by CWktart while the minimum intra-car load produced upon getting-off of the passengers at a given floor is represented by CWmkin with the intra-car load at the starting from the given floor being represented by CWks, the charged load CKkjn and the discharged load Cokut at the given floor are expressed by CWrzut(i)=CWstart(i) CWnnjn(i) (42) CWkin (i)=CWRks ( i)CWmkin(i) (43) where iare variables identifying the floors.

The above calculations are performed when the car k leaves the given floor (i) or in the doorclosed rest state of the car k.

Further, when the result of detection of the boarding and alighting passengers by a photoconductic device or ultrasonic sensor or the like is utilized only for lengthening the door time without detecting the going and coming of the passengers, the number of the boarding and the alighting passengers is set to O (zero) in preference over the results of the calculations in accordance with the expressions (42) and (43).Further, by counting the number of boarding and alighting passengers Nk(i), the average body-weight WPk(i) can be determined in accordance with the following expression (44):

Further, the weight of each passenger can be measured and collected in other traffic information table (corresponding to a table 256G 1 3 or 262A1 7). Further, on the basis of the signals produced by the photoelectric device or intra-car load detecting device, average boarding/alighting time (corresponding to table 262A6) required for each passenger to get on or off the car to thereby determine the speed at which the passenger gets on or off the car.

Additionally, the ratio of missing the car may be measured on the car basis. For example, the car missing ratio or the time involving the car missing for the car installed at the innermost part of the hall may be measured for the purpose of creating the optimal values for the door time not only on the floor basis but also on the car basis (i.e. for each of the floors and cars) to allow the automatic car-door control so that the car missing can be prevented while preventing the operation efficiency from being degraded due to the excessively long door open time.To this end, the time which has elapsed from the arrival of the car at the floor at which the hall call is issued to the detection of a first passenger or the probability (number of times) of the passenger not being detected with the door closing motion being triggered is detected to be collected in a table corresponding to the one 262Z4.

On the basis of the average value as determined, not only the door time but also the time limit for permitting the closing button to be turned off upon service for a hall call are prepared to be utilized in the elevator operation control.

The traffic information or data mentioned just above concerns the structure of the elevator car and the car disposition on the landing and little importance is put on the fine classification of this traffic information. Accordingly, this data may be collected and utilized in common to a plurality of characteristic modes.

Subsequently, at a step 2B2, collection of data for performance evaluation of the group supervisory control elevator system. This step is executed to confirm whether the algorithm and the parameters commanded by the learning microcomputer system have been optimal. The result of this evaluation is recorded. Beside, data for evaluating the operation performance is required for improving the simulation program itself in conformance with the actual situation.

For the purpose mentioned above, information corresponding to those of the tables 256G3, 256G4, 256G5, 262A10, 262A11,262A14, 262A15 and 262A20 are also collected.

These data can be detected from the values contained in the table provided internally of the operation control program 220 and the elevator control data table 210.

For example, data-of the continuation time of a hall call may be determined through procedures mentioned below.

At first, the hall call reset in response to the arrival of a car is detected and the corresponding value of a hall call continuation time table 21 0E is recorded in a collecting table. To this end, an accumulated value of the hall call continuation times and the number of times at which the hall call in concern is generated may be determined, by way of example. In this case, arrangement has to be made such that the hall call be reset in precedence to the resetting of the hall call continuation time.

Next, at a step 2B4, data required for allowing the control program for an equivalent elevator system for simulation to grow up to a program having the functions equivalent to the actual elevator system are collected.

Description will be made in this respect by taking as example a specification according to which the speed of all the cars is rated at 180 m/min. To this end, a car-based rated speed table 262Z1 1 is prepared on the basis of the specification speed read out from the elevator control table 225A. In practice, however, there is a speed tolerance in the range of 1 75 to 1 85 m/min. Moreover, the rated speed undergoes changes as a function of time lapse differently from one to another car. Further, it is inevitable that the speed varies in dependence on the travelling directions.Such changes or variations are detected on the basis of the speed values contained in the elevator state table 21 0A to be used for correcting the contents of the car-based rated speed table 262Z1 2.

Such correction can be accomplished successfully according to the inventive procedure mentioned above when the rated speed is set at a lower level by the speed controller in order to meet the energy saving requirement.

Further, there are recorded the times required for a car to run between different floors, e.g.

between first and second floors, first and third floors, first and fourth floors, second and third floors third and fourth floors. Additionally, data concerning the starting time which elapses from the stand-by state to the state ready to start the running, the door opening time, the door closing time and the service floor are detected on the car basis to be used for correcting correspondingly the relevant data contained in the table 262Z.

To realize the procedure mentioned above, the steps 282 to 2B4 are constantly executed once per second (step 225J), and a next step 2B6 is taken at every elapse of two minutes. Form the current traffic demand data collected at the step 2B2 and a day-based characteristic pattern table 262Z3 prepared on the day basis through learning of the characteristic modes as well as the relevant time zones, the algorithm parameters and relevant control parameters for realizing the optimum operation program are read out. Details of this step 2B6 are illustrated in Fig. 41.

Next, the number of passengers is determined from the traffic demand data collected at the step 282 to see whether it exceeds about 100 men (step 2254). If the answer is "NO", it is decided whether thirty minutes have elapsed at a step 225H. In this way, in case either the traffic volume data corresponding to 1 00 passengers has been collected or thirty minutes have elapsed, data collected primarily at the steps 2B2 and 283 are stored in a sampling table at a step 225K.

Next, on the basis of the data thus sampled and held, the characteristic elements of the traffic demands are arithmetically determined through the procedure similar to the one illustrated in Fig. 5 (step 225L) and sequentially recorded in a table array of the structure similar to that shown in Fig. 1 2 (step 225M).

By the way, the total traffic volume is recorded in terms of two elements, i.e. the traffic volume in the up direction CU(t) and the traffic volume in the down direction CD(t). As the data of the floors congested with the passengers getting on and off, the floor data l513(t) and IR1~3(t) as well as the boarding and alighting traffic volumes on the respective floors q513(t) and q1~3(t) are recorded.

A step 225J is provided for the purpose of making it possible to process in appearance an operation program creating software SF2B shown in Fig. 42 in parallel with the present program 2B notwithstanding of the fact that the task for the former is ranked lower than the latter. More specifically, once the processing under consideration has been completed, the OS executes the software SF2B, and after lapse of one second the program 2B being interrupted is regained, whereupon the step 282 et seq. are executed again.

Fig. 40 shows a flow chart of a program 2C (Fig. 38) for collecting the traffic information on the characteristic-mode base. In the first place, to identify the most approximating characteristic mode on the basis of the characteristic elements determined (step 225L) and recorded (step 225M) immediately before the instant program 2C is activated, the calculations mentioned below are carried out.

The characteristic mode number can be determined in accordance with the expression (27). In the instant case, however, it is determined in accordance with the modified expressions (25) and (26) through the procedures mentioned below.

First, the floor most congested or concentrated with the boarding/alighting passengers is represented by vectors QS and OR expressed as follows:

where I represents floor and q represents the traffic volume on the floor in concern.

Since the characteristic mode registration table 262Z6 includes the same characteristic elements for the congested floors as those mentioned above, the characteristic mode identifying function JX(t, m) considered only in the terms of the congested or concentrated floor can be given by

where m represents the identification numbers of the registered characteristic modes, X assumes a value of S or R where S represents the floor congested with the boarding passengers while R represents the floor concentrated with alighting passengers, and I represents the significant floor number of 1 to 3, i.e. represent the congested floors by order in dependence on the degree of congestion or concentration.

Vectors Pm and Pt for the typical one of the registered characteristic modes m and a spatial point of the characteristic mode t samples at a time point t are the vector Px given by the following expression (49).

Accordingly, scalar quantity Pt,m between the two points are given by

The characteristic mode M(t) having a typical point at which the above defined value becomes minimum is determined in accordance with the expression (27). The relevant scalar quantity L(t) is given by L(t)=MiN(Pt.m) (52) mE1 to9,11,19 Namely, it applies valid that L(t)=Pt, M(t).

Next, at a step 255B, it is decided whether the value of scalar quantity L(t) is substantially equal to the value of scalar quantity Pt,m relevant to other characteristic mode or there is difference between these value which can not be neglected. If there is non-negligible difference, the data sampled at the step 225K are stored in the traffic information storage table 256 (Fig. 30) for the identified characteristic mode through cumulative addition or the like.

Problem arises when the characteristic mode can not definitely identified or discriminated. To deal with such situation, there are conceivable various procedures which can be adopted in the processing of the step 255C.

Here, the procedure mentioned below is adopted.

In case the number of the similar characteristic modes is within three, the relevant traffic volumes are distributed in accordance with magnitude of the scalar quantity defined by the expression (51).

When there are many similar characteristic modes or when the scalar quantity L(t) related to the identified characteristic mode exceeds a predetermined value, it is considered as a new characteristic mode, which is then stored at a spared area I or II which is not allotted to any of the registered characteristic modes. In this case, the characteristic elements of this sampled data together with the time point tare placed in a spare area I or II of the table 262Z6 (Fig. 33).

In this way, the provisional registration of characteristic mode is accomplished, wherein the sampled data are abandoned when the new characteristic mode registered provisionally is not duly registered as the proper characteristic mode at a step 2E (Fig. 8).

In this conjunction, it should however be mentioned that the provisionally registered characteristic mode is used as one of the characteristic modes only in connection with the succeeding identification of characteristic mode for the collection of information at least on that day (or over one week at the longest). The spared areas I and II are used in this sequence. When these spared areas are insufficient for the provisional registration, the tables allotted to the unused characteristic modes (the tables for M17, M14, M8 and M9 are often idle) are made use of.

The control described above brings about the advantages mentioned below: O The traffic demand of the accidentai nature whose re-occurrence can not be foreseen can be excluded from the objectives for the learning.

O The traffic information concerning the new and important (significant and frequent) characteristic mode which makes appearance on a sudden can be collected as of the first day.

Fig. 41 shows a flow chart of the operation control parameter commanding program 2B6 (Fig.

39) for the control mode commanding parameters to the operation control program 220 on the basis of the values contained in a characteristic-mode based parameter table 274 which in turn is prepared through an operation-control-program creating software 279 (SF2B) mentioned hereinafter on the basis of the traffic demand learned on the characteristic-mode basis. The operation-control-mode commanding parameters thus prepared are transmitted to the operation control microcomputer M (step 275E).

At first, at a step 275A, the characteristic mode for the operation control is identified. This identification differs from those effected in connection with creation of the characteristic mode and collection of the traffic information. The purpose of this identification is to identify immediately or beforehand the Operation mode that is optimal for the current traffic demand.

To this end, at the step 275F, time elements are determined on the basis of the day-based characteristic pattern table 262Z8 in which day-based change patterns of the past main characteristic modes are recorded, the year-based pattern recording table 262Z9 for recording the numbers of the day-based characteristic mode patterns over one year and several months and the schedule table 290.

At the step 275A, the discriminative identification mentioned above is effected in consideration of the results of the step 275F.

This is because several minutes are required for detecting the traffic volume corresponding to 100 men. Basically, the expressions (44) to (52) are used while taking into consideration a prediction time element DP, the fact that one round time of a car is on the order of 2 min, and a hysteresis element HP provided for giving preference to the identified characteristic mode to deal with the situation where the traffic demand changes at a short period.

Now, the identification procedure at the step 275A will be described in more detail. At first, current operation characteristic mode (t) is duly determined from the registered characteristic modes in consideration of the time elements or factors mentioned above through the processes similar to those explained hereinbefore in conjunction with the expressions (45) to (52).

The function OP(t, m) for identifying each characteristic mode is defined as follows: Pttm OP(t, m)= DP(t, m) HP(t, m) The function DP(t, m) for the prediction time factor is given by

where

In this conjunction, it should be noted that the time zones of the duly registered characteristic modes have been recorded in two sets, as is illustrated in Fig. 13. The factor k,6 is selected as follows: k16 > 1 As this value of k,6 is greater, the characteristic mode becomes easier to be extracted in the relevant time zone learned in the past.

The function HP(t, m) for the hysteresis element contributing to stabilization of the characteristic mode to be outputted is given by

where j(t) represents the identification number of the characteristic mode used for the current operation control command, and k,7 > 1.

The characteristic mode 0(t) for the operation control is determined from OP(t, (t))=MiN1OP(t m)i (57) mEOM where OM represents an assembly of the duly registered characteristic modes and indicates to the characteristic-mode based parameter table 274 that the parameter of the characteristic mode belonging to the assembly have already been determined.

Upon identification of the characteristic mode 0(t) of the traffic demand, there are used in combination not only the same sampled and held but also the traffic demand data destined only for the characteristic mode identification independent of the collection of the traffic information, as is illustrated in Figs. 10 and 11.

More specifically, the processing indicated in the block 133 in Fig. 10 is carried out every ten minutes, and ten sets of values of the traffic demands prevailed ten minutes ago are recorded as indicated by the block B4, where on the basis of the sum of these recorded values the fourteen elementary values are determined in accordance with the expressions (50), (45) and (46).

Besides, with an attempt to increase the detecting speed, not only the table values being currently detected are taken into consideration but also supplement mentioned below is added.

The traffic volume CU(t) in theup-direction is added with the number of the up-hall calls being currently registered or the number of the waiting passengers at hall and the intra-car load of the car for the service in the up-direction or the number of the car calls witch respective predetermined weighting factors being applied. Similarly, the down traffic volume CD(t) are added with the values of the detected premonition thereof in precedence to the detection of the number of boarding/alighting passengers, to identify discriminatively a new characteristic mode at an earlier time.

Next, at the step 27 SB, it is decided whether the characteristic mode 0(t) is a definite one in consideration of the result of realization of the expression (35).

In other words, in case the values of OP(t, m) other than for the characteristic mode 0(5) can be neglected, the parameters relevant to the thus identified characteristic mode (t) are placed in the command parameter table 276 at a step 275D.

Otherwise, the optimal parameters for command are selected from the control parameters common to a plurality of relevant characteristic modes at a step 27 so.

For example, in connection with the door time, the number of the boarding passengers H(m) on a floor in concern are determined through interpolation in dependence on the value of the function OP(t, m) on the travelling direction basis, whereby the door time parameter DTi (Fig. 37) can be determined from the product obtained by multiplying the measured value of the average boarding/alighting time per person.

Fig. 42 illustrates a main flow of the software SF2B for creating the operation program, which is executed through the task activated once in a day by the step 2H shown in Fig. 28.

In execution of this software, it is at first decided whether or not the learning of the traffic information for a characteristic mode in concern has been performed (270A). Unless the traffic information for the characteristic mode in concern has been collected, the destined-floor-based traffic volume table C(i, j, m) and the number of car intended for use of simulation are established on the basis of occurrence densities H(i, m) and C(i, m) of the boarding and alighting passengers (Fig. 33) included in new traffic demand data (at step 270B).By way of example, description will be made in conjunction with Fig. 43 for the traffic demand in the up direction in the office-going time zone which belongs to the characteristic mode M1. As is shown in Fig. 43, the passenger density H(i, m) which is one of the traffic demand data as learned is divided in the up direction and the down direction to prepare arrays of HU(i, m) and HD(i, m), respectively. For the alighting passengers, the alighting passenger density data arrays CU(i, m) and CD(i, m) in up and down directions, respectively, are prepared. These four arrays in total are laid, respectively, along sides of the destined-floor traffic volume table C(i, j, m) in the manner shown in Fig. 43.Among thirty passengers who get on the car at the underground floor B 1 (i=1), the numbers of the passengers who leave the car at particular floors, respectively, can be concretely seen from Fig. 43.

In other words, a sum of the destined floor traffic volumes C(i, j, m) along the abscissa in the range of j > i represents validly the boarding passenger density HU(i, m), while the corresponding sum along the ordinate represents validly the alighting passenger density CU(i, m).

This means reversely that the destined-floor traffic volume can be estimated from the data of the boarding/alighting passenger densities on each floor in the up direction. More specifically, since all of the alighting passengers in number of 2.5 who leave the car at the first floor have all boarded the car at the floor B1, the number of the passengers riding on the car upon arrival at the second floor is 27.5.

Since the number of the passengers alighting the car at the second floor is 1 2 while the number of the passengers boarded the car at the second floor is 60, the number of the passengers on the car upon leaving the second floor is 87.5 inclusive of the remaining number of the passengers who boarded the car at the floor B1.

Assuming here for simplification that the passengers will leave the car at a certain ratio (12/87.5) independent of the floors at which they got on the car, the number of passengers who got on at the floor B1 and who are expected to leave the car at a next floor can be determined as follows: 12 x(30-2.5)=3.77 (58) 87.5 By performing the calculation mentioned above successively, all the destined-floor traffic volumes can be determined. Expression (59) mentioned below is for calculating the destined-floor traffic volume C(i, j, m) in the up direction on the condition that i- > j.

number of alighting passengers at floorj in up direction C(i, j, m)= number of intra-car passengers on arrival at floorj in up-direction remaining number of passengers from floor

The above expression can be encoded straightforwardly through recursive programming.

Next, at a step 270C, it is decided whether or not the characteristic mode under consideration has significant characteristics in respect to the traffic flow, utilization frequency of the special call and the like.

When decision is made that no significant characteristics are present (i.e. "NO" resulted from the step 270C) because the number of the passengers as observed in up and down directions and on the floor basis is substantially uniform, simulation is performed in accordance with a standard operation or service control algorithm to prepare a corresponding operation or service control program at a step 270E.On the other hand, when the passengers from a particular floor occupy a major part of the traffic volume (concentration of the passengers from a particular floor) as in the case of the characteristic mode M1 (corresponding to the traffic in the office-going time zone), simulation is performed in accordance with three algorithms for the division-based express service, the ordinary service program and a program to automatically dispatch to the congestion-concentrated floor a number of cars depending on the degree of congestion and cause the car to stand-by in the door opened state (step 270D).

In another exemplary case where ratio between the up-traffic volume and the down-traffic volume exceeds 1.5 or so, simulation is performed for preparing operation or service control programs separately in up and down directions for obtaining the width of the landing or the weighting factor required for preparing the stop call evaluation value or term Tc of the hall call allotting evaluation function (step H50-5 in Fig. 29).

In still another exemplary case where the evaluation is required in a time zone in which a coriding limit command is set as the objective or target value or where the evaluation is required to be done in terms of the service completion time, simulation is performed in accordance with a plurality of S 10 15 20 25 30 35 40 45 all allotment algorithms, as described hereinbefore in conjunction with the simulation program 271 shown in Fig. 19.

At the next step 270F, sets of the control parameters (such as the parameter e illustrated in Fig.

21 and parameter 3, shown in Fig. 23) are established for the operation on service control programs on the basis of the established algorithms, to allow simulations to be executed (step 270G). The results of the simulations made for each case are re-arranged on the parameter basis to be recorded in the table 272 (step 270F).

When the step 270F has been completed for all the parameters (step 270H) and for all the algorithms (step 270K), the operation or service control parameters are determined to establish the algorithm optimal for the imposed objective or target and placed in the characteristic-mode based parameter table 274 (step 270L).

When the optimal simulation parameters have been determined, simulation is again executed~for the operation or service control in concern, to determine subordinate parameters such as the probability of to prepare finally the characteristic-mode based parameter table 24 corresponding to the characteristic mode in concern (step 270M).

The processings described above are carried out for all the characteristic modes (step 270N), whereupon the program under consideration comes to an end. By the way, the due registration as the characteristic mode for the operation or service control as described hereinbefore is effected at the end of the step 270M.

An exemplary flow of the simulation executing program for the steps 270G and 270M is illustrated in Fig. 44.

The simulation programs can be broadly divided in a car processing portion (step A70) which includes a part for performing operations of the car (hardware system) itself (e.g. program part for simulating the running operation, door opening/closing operation and the like) and a part for performing the controls equivalent to those by the car control microcomputers E1 to En for controlling reasonably the car service, a group supervisory control processing portion for performing the control equivalent to the microcomputer M1 for realizing the supervisory operation or service control with high efficiency (e.g. program portions for the hall-call allotment control, dispersed operation control, division-based express service control and the like) and a portion for executing simulations (e.g.

passenger occurrence simulating processing A50, processing for collecting statistical data for recording the results A80, and processings for continuing simulation for a predetermined period A90, A100). Accuracy of the results of the simulation depends on the program arrangement for the steps A50 to A70 which constitutes the heart of the simulation and values of parameters and constants utilized therefore.

Accordingly, program should be so prepared that the items mentioned above which exert great influence to the performance of the group supervisory control become equivalent as closely as possible to those of the actual elevator system on one hand, and on the other hand the constant and parameters as used should have as accurate values as possible.

Now referring to Fig, 44, the constants and the parameters mentioned above are at first modified or re-established at proper values by using the values of the information learning table 262 (Fig. 33) at the step Al 0. The constants are retrieved mainly from a common information table 262Z while the parameters are retrieved mainly from a table correspdnding to the simulation parameter table 262A22 in which concrete control parameters learned for each simulation parameters are recorded, and they are determined through selection and arithmetic operation based on the algorithm parameters and simulation parameters therefor supplied as for the activation conditioning.

In conjunction with a simulation case in concern, establishment of the constants and parameters relative to the invariable car speed, door opening/closing time, traffic volume and the like is carried out in the manner similar to the case where establishment of traffic information 270B shown in Fig. 42 or the like are performed, so as to decrease duplication of processings.

Further, generation of the passengers through generation of random numbers can be processed at high speed by preparing previously a passenger occurrence (or generation) table containing data of time point t and fioor i andj at a step 270B.

Next, initialization of simulation variables is performed (step A20), which includes initializations of a random number of the passenger generating processing and the hall call table. At a step A30, initialization of variables for the statistical processing and the like are effected. In this case, a statistic table is initially established. At a step A40, time is set to zero and incremented to a predetermined value at step A90 (in this case, time is incremented by one.) At a step Al 00, it is decided whether time lapse exceeds the predetermined value mentioned above. The processing including the steps A50 to A90 is performed until the time lapse exceeds the predetermined time.At a step A50, the passenger generation (occurrence) processing is performed, at a step A60, the group supervisory processing for allotting the hall calls, if present, is performed, at a step A70, the car processing relative to the running and stop of the car, door opening/closing and the like. At a step A80, a statistic data collecting processing is carried out.

Here, description will be made concerning the processing including the steps A50 to A70. The passenger generation (or occurrence) processing at the step A50 is performed on the basis of the predicted data of the destination traffic volume obtained through execution of the program 270B for establishing the traffic information for simulation, to determine the number of the generated passengers migrating from the floor ito the floorj by using a uniform random number to thereby cause a hall call at the floor i. Subsequently, through the group supervisory control processing at the step A60, allotment of the above hall, if general, is effected. The method of allotting the call is same as the one described herein before in conjunction with the operation or service control program.Through the car processing at the step A70, processings for the car running and stop states, opening/closing of door, generation of car call and the like are performed.

Next, description will be made on the statistic data collection processing and the statistic processing at the steps A80 and Al 10 by referring to a flow chart shown in Fig. 45. At steps A80-1 to A80-4, the number of loops are determined for the car directionj, siniulation parameter e, traffic demand division m and the floor i, while at steps A80-6 to A80-8, decision has been made whether the loop fore, m and i have been completed. At a step A80-5, the statistic data (number of car stops, number of hall calls, number of boarding passengers, number of alighting passengers and the like) are collected for each of e, m and i.

Figs. 46 to 49 show flow charts for illustrating the subordinate parameters such as for probability of stop, prediction of full-car or saturated state, preference level of service, control of door time and the like as an example of the statistic processing program used at the steps A80-5. Referring to Fig. 6 which illustrates in a flow chart a program for determining the stop probability, the number of loops for the floor i is established at a step BP 10. On the basis of a number of reversions of car travelling direction d, the number of floor-based hall calls f, and the number of car calls, the number of car stops is determined, from which the stop probability PS, at floor i (in each direction) is determined.This probability PS, can be determined in accordance with the following expression: f,(=number of car stops at floor i in directionJO+g' PS,=a1, (60) number of reversions in direction d In this connection, it is preferred from the viewpoint of the processing time that the data d, f, and g1 be determined at the step A80, while the parameter PS; be determined at the step Al 10. By the way, a11 is a predetermined factor. This factor a11 is usually set at 1 (one). If the results of simulation are out of conformity with the actual car, the value of this parameter is correspondingly corrected to bring about the conformity.

Referring to Fig. 47 which shows in a flow chart a full-car prediction program, the number of loops for floor i is established at a step BF1 0. On the basis of the number of the hall calls fj at the floor i retrieved from the hall-call number table 272A6 and the number of the car boarding passengers H, from the floor as retrieved from the car boarding passenger density table 262A7, the full-car prediction value Phi for the hall call at the floor i (average number of boarding passenger in a single hall-call service) is determined (step BF30).Further, on the basis of the number of car calls g1 to the floor i retrieved from the car-call number table 272A7 and the number of alighting passengers H1 at the floor i read out from the car alighting passenger density table 262A7, the full-car prediction value for the car call to the floor i is arithmetically determined (step BF40).The full-car prediction value Phi for the hall call and the full-car prediction value Pci for the car call are, respectively, determined as follows: TS boarding passenger density H, from floor i Ph1=a12 - (61) number of hall calls f, at floor TS alighting passenger density C, at floor i PC,=CX 13 (62) number of car calls g1 to floor i where TS represents the time required for simulation, and 12 and a13 are predetermined factors or coefficients, respectively.

Referring to Fig. 48 which shows a door time control parameter determining program. At a step BD10, the number of loops for the floor i set. At a step BD30, the door time control parameter TD is arithmetically determined from the number of hall calls fi at floor of the hall-call number table 272A6, the boarding passenger density H, generated at floor i read out from the car boarding passenger occurrence density table 262A7, the number of car calls 9 to the floor i read out from the car-call number table 272A7 and the alighting passenger occurrence density C1 at the floor i read out from the car alighting passenger occurrence density table 262A6.The door time control parameter in concern can be expressed as follows: Number of boarding passengers at floor i (i.e. TS H,) TD1=a15 number of hall calls fi at floor i number of alighting passengers at floor i (i.e. TS Hj) +(g16 (63) number of car calls 9 to floor i (63) where sag15 and a16 are predetermined factors.

The parameter TD, resulted from the above calculation is recorded on other table 272A1 1.

Alternatively, the door time control parameter TD, for a characteristic mode in concern may be calculated in the earlier half of the step 275E shown in Fig. 40 and directly placed in the table 276G.

The latter mentioned method allows the table for recording the results of simulation to be realized in a reduced size.

The statistic processings are effected in the manner described above, wherein the various values resulted from the processing are made use of in the operation control programs shown in Figs. 26 to 29.

Further, these values are used at the step A10 or step B as the constants or parameters in the program for the group supervisory processing at the step A60 and in the program for the car processing at the step A70. In consideration of the number of simulations, arrangement is made such that these values can be used for the characteristic mode m in concern on the next day.

Fig. 49 shows in flow chart a program for calculating the subordinate parameters and determining the effect of the operation control program created for each of the characteristic mode (i.e.

on the characteristic mode basis or base). This program is to illustrate details of the step 270M shown in Fig. 42.

At first, the simulation parameters optimal for the objectives as determined at the step 270L are set (step 270M1), to perform again simulation (step 270M2). Through the statistical processings (steps A80 and Al 10) illustrated in detail in Figs. 45 to 48, the subordinate parameters are prepared by using the optimal simulation parameters and recorded in the characteristic-mode based parameter table 274 (step 270M4).

Next, for the traffic demand belonging to the characteristic mode in concern, simulating conditions for the operation control program delivered by the manufacturer are set (step 27on4), to perform the simulation (step 27ops). In other words, specifications of the program portions for performing the group supervisory processing A60 and the processings equivalent to the car control microcomputers for car processings and determined on the basis of the operation control parameters contained in the elevator control specification table 225, to thereby perform the simulation.

On the basis of the results of the simulation as performed, the delivered program and the operation program created optimally for the given objectives are compared with each other in terms of the evaluation values, the results of which are recorded (step 270M6).

In the comparison mentioned above, one or more items are compared with regard to the evaluation indexes for serviceability (waiting time of hall call, continuation time of hall call at each floor in each direction, continuation time of car call, service completion time, distribution of these times or average thereof and long waiting times each having a by-5% lengthened waiting duration, ratio of reservation change and first arrival ratio) and power saving properties (overall power consumption coefficient determined from the ratio between the run and the stop of each car and the car-based power consumption factors).

As a modification of the step 27on7, the step 270B may be included in the step 270M7 as a part thereof and the factors or coefficients 11 to ,6 are correctively modified in dependence on the results of the comparison mentioned above.

Next, in order to command the parameters which can assure the optimal operation control program even when the number of the cars is changed due to special operations and maintenance, simulation in which changes in the number of the cars are taken into consideration is performed, the results of the simulation also being recorded.

Fig. 50 shows in flow chart for illustrating an example for calculating the optimal simulation parameters in dependence of the number of elevator cars in service on the assumption the number of the installed cars is six. In the first place, it is decided whether a characteristic mode in concern is the one (typical or representative mode) in which the product of the utilization frequency and the traffic volume is at maximum (step SC 1). If not so, simulation is performed on the basis of the average number of cars operated in the characteristic mode in concern on the last day (step SC3), while in the ordinary traffic pattern the simulation is performed with six or four cars (step SC2). It goes however without saying that simulation with six cars may be omitted at the steps SC3 and SC2.Subsequently, on the basis of the results of two simulations effected with six and four cars, respectively, the optimal simulation parameters for the operations with five or three cars, respectively, are arithmetically determined through interpolation (step SC4).

In the case of the example mentioned above, two different numbers of the cars are employed. It is however possible to calculate the optimal simulation parameters for the Operation with the remaining numbers of the cars even when three or more different car numbers are employed. As is obvious, the greater the different car number is, the more accurate results of simulation can be obtained. Although the selected car numbers are assumed to be four and six in the illustration of the flow chart, any numbers may be selected in the range of the number of the installed cars.

Here, description will be made on the interpolation mentioned above.

Since the optimal simulation parameters which contribute minimization of the average waiting time may safely be regarded to be proportional to the number of cars, there is-conceivable proportional interpolation method. As method which assures higher accuracy, interpolation may be effected in proportion to 1.6-th power of the car number. Further, predetermined factors may be involved in the interpolation in dependence on the traffic volume.

Besides, arrangement may be made such that the coefficients or factors such as 1.6-th power and the like are calculated beforehand through simulation performed by taking advantage of the idle time zone.

Fig. 51 graphically shows example of the average-waiting-time curves prepared through a predetermined interpolation on the basis of the contents of the curve data table 272A1 obtained through simulation. The interpolation as adopted is a well inown interpolating method in which a quadratic curve approximation is made with three pieces of adjacent data.

The curve f4 corresponds to the case where the car in service is four in number, while the curve f6 corresponds to the case where the car in service is six in number. It is shown that the simulation parameter at which the average waiting time of the case represented by the curve f4 is minimum is a4, while the simulation parameter at which the average waiting time of the case represented by the curve foe is minimum is a!6. On these conditions, the proportional interpolation method is given by the following expressions: a6=a6+(a4+a5)/2 (64) a3=a4+(a4+a6)/2 (65) where cg5 and a3 represent the optimal simulation parameters in the cases where the car numbers are 5 and 3, respectively.

In the foregoing, the second exemplary embodiment of the present invention in which the microcomputers M and M2 are used has been described so far as the structure or arrangement is concerned. In the following, operations of this exemplary embodiment will be supplementarily elucidated with the aid of Figs. 52 to 57.

Fig. 52 is a view for graphically illustrating a typical example of the traffic demand likely to take place in the lunch time zone in a building occupied by a single company. Variations in the overall traffic volume as a function of time is shown at (a).

There are shown at (b) in Fig. 52 a curve CM2 representative of changes in magnitude of the traffic flow belonging to a characteristic mode M2 (earlier half of lunch time) in which the passengers going to seventh floor which is a dining hall floor (i.e. alighting passengers at the seventh floor) occupy a major part of the traffic volume and a curve CM2 which represents changes in magnitude of the traffic volume belonging to a characteristic mode M3 (later half of the lunch time) in which the passengers returning to the respective starting floors from the seventh floor (i.e. boarding passengers at seventh floor) occupy a major proportion of the traffic volume. Figs. 53a and 53b show distributions of boarding passengers on the floor (direction) base in both the above mentioned characteristic modes M2 and M3, respectively.

Problem arises in the time zone where the curves CM2 and CM3 intersect each other. In a time zone from t32 to t33 in particular, the traffic volume is moderate. It will certainly be possible to establish separately a characteristic mode inherent to this time zone. However, in consideration of the fact that the sum of the passengers in number determined by the product of the duration of this time zone and the traffic volume is small relative to the registered characteristic mode and that the characteristic mode in concern may not be considered as the significant one because the characteristic modes M2 and M3 having similar behaviors have been established, the characteristic mode in concern which may be generally referred to as the mid lunch time mode will be unnecessary. (However, in the case of hotels, rental office-building or the like, the mid lunch time traffic mode may be profitably established.

By the way, establishment of this characteristic traffic mode is controlled by the characteristic mode creating and identifying program shown in Fig. 38).

Such being the circumstance, the traffic mode in the time zone t32-t33 is decided to be a swilight characteristic mode relative to a plurality of other characteristic modes as described hereinbefore in conjunction with the step 255B shown in Fig. 40 and the step 275B shown in Fig. 41 to thereby prevent the number of the characteristic modes from being increased indefinitely while assuring the collection of the proper traffic information and appropriate control of the command of the operation control parameters. An apparatus imparted with functions for reporting and monitoring these data programs is shown in Fig. 63.

Next, referring to Figs. 54 to 56, description will be made of examples of reports issued automatically by the learning microcomputer system and a manner in which the prediction of occurrence of the characteristic modes can be made with an improved accuracy as a function of the calendar day and time at the step 275F (Fig. 41).

Fig. 54 shows an example of the report on contents of the day-based characteristic pattern table 262Z8 shown in Fig. 34, which report is displayed on the CRT of the terminal device PD or printed out through a printer. The characteristic modes are displayed as re-arrayed in the order of P1 to P 1 in dependence on magnitude of stress or burden (e.g. traffic volume) imposed to the elevator control system. From the chart shown in this figure, difference in traffic between the office-attending day DP3 (day-based pattern curve fop2) and the holiday such as Saturday, Sunday, festival day or the like (daybased pattern curve fDp4) can be readily understood at a glance, to an advantage.

Fig. 55 shows an example of a report on the contents of the over-year recording table, while Fig.

56 shows a display of the day-based pattern prepared through prediction for the instant year.

This prediction is carried out in the manner mentioned below.

It should first be mentioned that the highest priority or preference is put on the alternation of holidays, setting of special holidays, a day-based traffic pattern for a day on which an event is held as well as the relevant date and the like which are inputted with the aid of a keyboard of the terminal device PD as the sheduled reservation, as occasion required.

Next, decision is made as to whether it is a festival, which is followed by the decision as to whether it is day on which identical day-based traffic pattern is expected every year, as is the case of a national holiday or a fete-day in commemoration of the founding of company (refer to Note 1 in Figs.

55 and 56).

Subsequently, conversion of the days of the week is effected in dependence on year, and the daybased traffic patterns for the days of the week are determined through the statistical processing, to prepare the predicted day-based patterns.

Through the prediction of the day-based patterns mentioned above and the learning of the characteristic modes for every day-based pattern as well as the relevant time zones, prediction of the characteristic modes over an elongated time span can be accomplished with higher accuracy than the processing described before in conjunction with the step 275F.

In connection with the step 2F (shown in Fig. 38), the additional establishment of the day-based pattern and the learning of the time zones are carried out such that no consideration is paid to the presence or absence of the characteristic mode of duration on the order of 10 minutes or to the difference between the adjacent characteristic modes appearing every day, but the time zones of the characteristic modes coinciding with each other are learned.

Fig. 57 shows an example of a report on the results of the learning system which can perform the learning function in such a broad sense that it can be encompassed by the concept of the so-called artificial intelligence control for controlling the creation of the characteristic modes, learning ofthe traffic information and the creation of the operation control program in a systematic manner.

The learning process is easier to be understood by tabulating or drawing charts for each of the day-based patterns.

By the way, the phrase "sudden change" is used to encompass not only the newly added characteristic modes but also a remarkable change in the configuration of the day-based pattern DP3.

Further, the contents of the format RDP3 for reporting the results of learning are prepared on the basis of the evaluation data table (e.g. frequency distribution table 262A10 for continuation time of hall call) constituting a part of the information learning table 262, the results of simulation performed at the steps 27on4 and 27on5 (Fig. 49) and the time zone data of the day-based pattern table so that the evaluation values corresponding to one day are reported in terms of the average values for ten hours in the daytime.

In this manner, it is possible to confirm that the learning system operates in order, while the effect due to the equipment of the learning system can be quantitatively detected, to further advantages.

In the foregoing, the second embodiment of the invention has been fully described.

In the following, other embodiments of the invention as well as modification will be added.

At first, modifications of the characteristic mode and the day-based pattern will be described by referring to Figs. 58 and 59.

Fig. 58 shows a modification of the embodiment shown in Fig. 1 9, while Fig. 59 shows a version of those shown in Figs. 33 and 36.

The main characteristic feature of the modifications resides in that the characteristic modes are discriminated on the day basis. For example, the learning of the traffic information in the office-going time zone is performed in conjunction with the office-going zone table for the day in concern after determination of the day-based pattern through execution of the traffic demand characteristic determining program 257.

In this connection, it is preferred that the first traffic demand characteristic determining program 254 be so prepared that the time zones can be identified on the day basis, because then the program can be realized in a much simplified form. Additionally, arrangement may be made such that the time zone data in concern is inputted through the terminal device PD, as occasion requires, to the similar effect.

The method of identification on the day basis may be carried out in such a manner that prediction is first prepared in dependence on the traffic volume in the office-going time zone and finally decision is made on the basis of the data concerning the total number of users in a day. In this way, the program 257 also can be realized in a simplified structure.

Additionally, there is conceivable a method of determining the first traffic demand characteristic according to which the evaluation function is prepared for each of the characteristics (in common to individual days, if desired), wherein the characteristic for which the evaluation function assumes the greatest value is identified.The evaluation function XUP in the office-going time zone is expressed by ÇUp=k2, number of boarding passengers from lobby floor/(number of boarding passengers from dining and non-specific floors+number of alighting passengers on lobby floor) (66) In similar manners, the characteristic evaluation functions FLTA5LTB and 6Dp in the earlier half of lunch time, the later half thereof and in the office-leaving time zone, respectively, are given by number of alighting passengers at dining floor BLTA=k22' (67) number of alighting passenger on non-specific floors number of boarding passengers from dining floor LTB=k2i (68) number of boarding passenger from non-specific floors number of alighting passengers on lobby floor fDP=k24 (69) (number of alighting passengers on non-specific and dining floors+number of boarding passengers from lobby floor) Discrimination among the ordinary traffic, the ordinary congestion and thin traffic can be made simply in terms of the traffic volume. In the thin traffic time zone, learning of the traffic information is rendered unnecessary because of small traffic volume.

The characteristic evaluation function XBT for the ordinary traffic and the ordinary congestion is given by traffic volume on non-specific floors fBrk=5 . (70) traffic volume on lobby+ traffic volume on dining floor In conjunction with the expressions (67) to (70), k2, to k25 represent weighting factors, respectively, where k22 and k23 are 4, respectively, k25 is 2, and k2, and k24 are 1, respectively.

As the decision output, the characteristic for which the values calculated in accordance with the expressions (66) to (70) are maximum is selected.

Next, other modifications of the embodiment shown in Fig. 1 9 will be mentioned. Fig. 60 illustrates supplementarily the flow of data of the learning system in the software arrangement shown in Fig. 19.

In conjunction with the second embodiment, it has been described that the basic operation program is created by executing beforehand the simulation program 271. In this connection, modification is made such that an additional simulation for emergency may be performed in conjunction with execution of the control parameter preparing program 277 when significant change such as, for example, alteration in the desired or target values as set and/or change in the number of cars for service does occur. In particular, simulation for calculating the subordinate parameters at the step 270M2 (Fig. 49) can be again performed only once, allowing the command parameters to be outputted relatively speedily. The characteristic-mode-based parameter table 274 can then be realized in a correspondingly reduced size, to an advantage.

In conjunction with the program A50 shown in Fig. 29, there will be mentioned below an improved proposal for giving preference to the service for the floor where the number of passengers awaiting car at hall is great, with a view to minimizing the average waiting time of the individual passengers.

Fig. 62 shows in a flow chart the concept of the above proposal. The heart of the improvement resides in the manner in which the predicted waiting time TS is prepared at the step H50-2.

More specifically, of the floor where the frequency of occurrence of passenger is extremely low, the value of the predicted waiting time is selected equal to the value of TS. On the other hand, in the floor where the frequency of occurrence of passenger is high or moderate, the evaluation value is selected greater than the predicted waiting time in dependence on the degree of utilization.

In the foregoing, description has been made in conjunction with the examples shown in the drawings. In the following, other examples of modification and application will be briefed.

A part of the learning system such as, for example, the operation program creating block 270 for the simulation program 271 may be implemented in a transportable independent device, which is then connected to the learning microcomputer system through an adapter once for a month to prepare the optimal operation program through repeated simulations performed on the basis of the general traffic information or leaning on the character-mode basis, wherein the created operation program is then recorded in the learning microcomputer in terms of the various relevant parameters.

In this case, the program creating device of portable type mentioned above may be periodically connected to a building supervising apparatus or an elevator supervisory control apparatus for data communication, to thereby prepare the operation program automatically.

The foregoing description has been made in conjunction with the group supervisory control.

However, this is only for the purpose of illustration. The invention can equally be applied to a system where a single car is operated. It goes without saying that the invention can be applied also the port type elevator system which has long been known and in which the destination fidor can be designated at the landing. In that case, the destination call button corresponds to what is called the hall call button herein. In other words, the hall call as referred to by the invention may be generated by any types of apparatus installed at the landing. The invention makes it possible to realize the elevator control which can promptly adapt the operation of the elevator system to the current traffic demand in a building.

Further, the creation of characteristic modes, collection of the traffic demand information for each of the characteristic mode, creation of the operation program and others described in conjunction with the embodiments of the invention can be separately or independently made use of without impairing the respective functions.

The exemplary embodiments of the invention have been described in connection with the elevator control. However, the detection of the traffic demands in a building is useful in supervising the building. Accordingly, the invention can also be applied to a building supervising system for controlling the air condition in consideration of the flows of persons and the number of residents and for preventing unauthorized invasion.

According to the invention, the elevator control can be performed in conformance to the traffic demand in a building to assure the elevator service with an improved efficiency.

Claims (27)

Claims

1. A group supervisory control apparatus for an elevator system, comprising a plurality of elevator cars operated for affording service to a plurality of landings of a multi-floor structure and means for detecting traffic demand in the elevator system, said plurality of elevator cars being subjected to supervisory control as a group through control means including variable elements, further comprising: means for preparing and storing said variable element for each of plural characteristic modes which represent identificably a corresponding number of discriminatable states of the traffic demand; and means for identifying the characteristic mode to which the detected traffic demand state belongs and supplying as command to said control means the variable element corresponding to said identified characteristic mode.

2. A group supervisory control apparatus for an elevator system according to claim 1, said variable element including an operation control program for the group supervisory control or variable parameters for said group supervisory control.

3. A group supervisory control apparatus for an elevator system according to claim 1 , further including means for collecting traffic demand information for each of said characteristic modes, wherein said variable element storing means includes means for creating a variable element corresponding to said collected traffic information, and means for storing the created variable element.

4. A group supervisory control apparatus for an elevator system according to claim 1, including means for establishing the characteristic modes of the traffic demand for the control of said elevator system and means for creating said control-destined characteristic mode in dependence on said detected traffic demand.

5. A group supervisory control apparatus for an elevator system according to claim 4, said characteristic mode establishing means is so arranged that a plurality of the characteristic modes can be previously established in an arbitrary manner.

6. A group supervisory control apparatus for an elevator system according to claim 4, wherein said creating means includes means for extracting characteristic mode from said detected traffic demand.

7. A group supervisory control apparatus for an elevator system according to claim 6, wherein said characteristic mode establishing means is so arranged as to establish the control-destined characteristic mode on the basis of the characteristic mode extracted by said extracting means.

8. A group supervisory control apparatus for an elevator system according to claim 6, wherein said creating means is so arranged as to create the control-destined characteristic mode corresponding to said extracted characteristic mode extracted by said extracting means.

9. A group supervisory control apparatus for an elevator system according to claim 6, wherein said creating means is so arrarrged as to establish a new control characteristic mode on the basis of said extracted mode unless said extracted mode corresponds to any one of said control-destined characteristic modes.

10. A group supervisory control apparatus for an elevator system according to claim 4, wherein said creating means includes means for recognizing characteristic data provided by said detected traffic demand, means for recording said recognized characteristic data, and means for extracting the characteristic mode of the traffic demand of a predetermined duration from said recorded characteristic data.

11. A group supervisory control apparatus for an elevator system according to claim 4, wherein said traffic demand detecting means includes means for storing the traffic information data of said elevator system, and means for outputting the stored data when the stored data of predetermined information has attained a predetermined amount.

12. A group supervisory control apparatus for an elevator system according to claim 4, wherein said traffic demand detecting means includes data storing means for storing traffic information data of said elevator system, storage time measuring means for measuring the storage time, a traffic demand detecting timing generating means for determining the period for detecting said traffic demand from at least said stored data, and detected data outputting means for determining the density of the traffic demand from said stored data and the storage time to thereby output said density as part of the traffic demand detection data.

13. A group supervisory control apparatus for an elevator system according to claim 6, wherein said creating means is so arranged as to decompose said detected traffic demand into a plurality of basic elementary components, to thereby express characteristic of the traffic demand on the basis of elementary function values representing magnitudes of said elementary components, respectively.

14. A group supervisory control apparatus for an elevator system according to claim 10, wherein said recognizing means includes means for recognizing at least an element representing the whole traffic volume, and means for recognizing elements representing migrations of passengers among the floors, whereby the characteristic of said traffic demand is recognized on the basis of said plural elements.

1 5. A group supervisory control apparatus for an elevator system according to claim 14, wherein said means for recognizing the element representing the migration state of the passengers determines distribution function of utilization frequency on the floor base for establishing-the value of said element.

1 6. A group supervisory control apparatus for an elevator system according to claim 14, wherein means for recognizing the element representing the migration state of the passengers detects the landing having higher frequency of utilization than the other landings, to set a floor number for identifying said detected floor as the value of said element.

17. A group supervisory control apparatus for an elevator system according to claim 10, wherein said traffic volume detecting means is so arranged as to produce the output at least when the predetermined traffic volume is exceeded, while said characteristic recognizing means is so arranged that said characteristic data recording means records as new characteristic recognition data said characteristic recognition data of the traffic demand outputted by said traffic flow detecting means, to thereby vary the characteristic recording period in dependence on the density of the traffic demand.

1 8. A group supervisory control apparatus for an elevator system according to claim 10, wherein said recognized data recording means is so arranged as to perform the recording for a predetermined time, to thereby allow the characteristic mode of the traffic demand to be continuously created even after the establishment of the characteristic mode.

19. A group supervisory control apparatus for an elevator system according to claim 10, wherein said recognized data recording means is so arranged as to record the period or time of the recognized traffic demand in corresponding combination with the recognized data.

20. A group supervisory control apparatus for an elevator system according to claim 1 0, wherein said recognized data recording means is arranged so as not to record the recognized data of the traffic demand, in case said data representing the characteristic recognized by said characteristic recognition means is insufficient to be extracted by said characteristic extracting means.

21. A group supervisory control apparatus for an elevator system according to claim 10, wherein said characteristic mode extracting means is so arranged as to extract the characteristic mode included in the traffic demand in the elevator system after the recognized data of traffic demand of a predetermined duration has been recorded by said characteristic data recording means, to thereby extract as the characteristic mode the characteristic occurring at a high frequency.

22. A group supervisory control apparatus for an elevator system according to claim 21, wherein said characteristic mode extracting means determines on the characteristic base a sum of time of the traffic demand states which have characteristics coinciding with or similar to the characteristics exhibited by said recorded characteristic data, to thereby extract the characteristic having duration longer than a predetermined time limit, whereby typical characteristic are extracted among a lot of characteristics exhibited by the traffic demand in the elevator system.

23. A group supervisory control apparatus for an elevator system according to claim 20, wherein said characteristic mode extracting means determines on the characteristic base an evaluation function in accordance with a characteristic mode extracting evaluation formula composed of a plurality of elementary functions, whereby a predetermined number of characteristic modes are extracted from the values of said characteristic mode extracting evaluation function.

24. A group supervisory control apparatus for an elevator system according to claim 23, wherein said characteristic extracting evaluation formula includes at least function terms concerning degree of significance of the characteristic, the time for which said characteristic makes appearance and the continuity of said characteristic.

25. A group supervisory control apparatus for an elevator system according to claim 7, wherein said creating means is so arranged as to establish as said control-destined characteristic mode the mode extracted by said extracting means continuously for a predetermined period.

26. A group supervisory control apparatus for an elevator system according to claim 6, wherein said creating means is so arranged as to evaluate said control-destined characteristic mode and said extracted mode through comparison to thereby alter said control-destined characteristic mode.

27. A group supervisory control apparatus for an elevator system according to claim 5, wherein said characteristic mode setting means includes an external setting device for mode setting.