JPH0764488B2 - Elevator group management control device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to an elevator group management control device effective for assigning a hall call to an optimal elevator when a plurality of elevators are installed in parallel. It is about improvement.

[Conventional technology]

As a typical method of elevator group management control, there is a call allocation method that selects an optimal car using an evaluation function each time a hall call occurs and allocates that call. The problem of which car is best assigned is a combinatorial optimization problem including statistical elements, and it is generally impossible to obtain a solution by an algorithm (mathematical method).

Therefore, in the current elevator group management system, some "common sense rules" that have been judged to be effective based on many years of experience, such as "the predicted travel time to a newly registered hall call is the shortest Select the car first. "
Alternatively, use an empirical rule such as “If you assign a new landing call, avoid allocating it to a car in which other calls that have already been assigned will wait long”. In the case of incorporation, or in the assignment method using the expert system, the assignment is performed by incorporating it in the knowledge database.

In order to solve such a problem, it is possible to automatically learn and generate a decision system that determines the optimal assigned car by using a neural net, which is completely different from the conventional method. A new group management control device for elevators has been proposed (for example, Japanese Patent Application No. 63-
No. 105633 "Elevator group management control device"), and its outline will be described with reference to FIGS. 2 to 6. First, the neural network will be briefly described.

A neural network is a network that imitates the human brain, in which multiple units corresponding to neurons (nerve cells) in the brain are connected intricately, and the operation of each unit and the connection form between units are described. By deciding well, you can embed the pattern recognition function and the knowledge processing function,
For example, it is introduced in "Nikkei Electronics" August 10, 1987 issue (No427), P115 to P124.

First, FIG. 2 shows the structure of a unit that models a neuron. The unit U _i transforms the sum of the inputs Q _j from other units by a certain rule to obtain Q _i , and the joints with other units have variable weights W _ij respectively. This weight is for expressing the strength of the coupling between each unit, and if this value is changed, the structure of the network will be substantially changed without changing the connection. The learning of the network described later is to change this value, and the weight W _ij takes positive, zero, and negative values. Zero means no bond.

When a unit receives inputs from multiple units, if the sum of the inputs is represented by NET, the sum of the inputs of unit U _i is Is.

Each unit applies the sum NET of this input to the function f and converts it into the output Q _{i as} shown in the following equation.

This function f may be different for each unit, but generally, the threshold function shown in FIG. 3 (a) or the sigmoid function shown in FIG. 3 (b) is used.

This sigmoid function is a differentiable pseudo-linear function, Can be expressed as The value range is 0 to 1, and approaches 1 as the input value increases and approaches 0 as the input value decreases. When the input is 0
It becomes 0.5. Add a threshold θ (bias), In some cases.

FIG. 4 is a diagram showing an example of the structure of the network, and the illustration of the weight of the coupling portion between the units is omitted.

The neural network is classified into a pattern associative type and an automatic associative type according to the structure of the network. Here, the pattern associative type will be used for description. The pattern associative type is a network for converting an input pattern into a certain output pattern, and each unit is hierarchized into an input layer, an intermediate layer, and an output layer as shown in FIG. Each unit is connected from the input layer to the output layer, but the units in each layer are not connected. Moreover, the input unit and the output unit are independent.

In such a neural net, when input data is given to each unit of the input layer, this signal is converted by each unit, transmitted to the intermediate layer, and finally comes out from the output layer. It is necessary to set the strength of the coupling between the units, that is, the weight, to an appropriate value. The setting of the weight is performed by learning the network as follows.

First of all, set all weights randomly,
Input data for learning (data in which desired output is known in advance) is given to each unit in the input layer. Then, at this time, the output value output from each unit of the output layer is compared with the desired output value, and the value of each weight is corrected so as to reduce the difference (error). Then, this is repeated using a large number of learning data until the error converges. A learning algorithm for correcting the value of each weight from the value of this error will be described later.

When the learning is completed in this way, the knowledge processing device is automatically embedded in the neural network, and the desired output can always be obtained not only for the learning data but also for the unknown input data.

Note that this neural net can be configured as an LSI by configuring each unit using amplifiers and resistors, but it is also possible to configure a virtual neural net with software and process all by arithmetic operations. It can be realized by a computer.

FIG. 5 is a diagram showing one embodiment of the overall configuration when this neural net is used for elevator allocation control. Here, for the sake of convenience of explanation, only two elevators, the first and second elevators to be controlled, are controlled. However, it is needless to say that the same configuration can be applied to any number of units.

In FIG. 5, 1 is a hall call button (1
Only one floor is shown, the others are omitted) 2 is a landing call signal, 3A is an operation control device that manages the operation of Unit 1,
Similarly, 3B is an operation control device that manages the operation of the second car, 4 is a car information signal indicating the state of each car (car position, direction, stop, running, door open / closed state, car call, load, etc.), and 5 is a group. A microcomputer for performing an allocation function as a management device, each input serving as an input to the neural network based on each data of the hall call signal 2 and the car information signal 4 read via the input / output interface 6. An input pattern calculating means 7 for calculating pattern elements, a neural net 8, and an allocation judging means 9 for judging which machine is optimum from the output of the neural net 8 are provided, and the allocation result is input / output interface. It is output as an allocation signal 10 via 6. Each operation control device 3A and 3B controls the operation of the car so as to sequentially respond to the landing call assigned by the assignment signal 10 and the car call registered in its own machine.

FIG. 6 is a diagram showing an example of each input pattern element input to the neural net. A _{1 to} d ₁ are input pattern elements representing the situation of the No. 1 machine, and a _{2 to} d ₂ are the input pattern elements of the No. 2 machine, respectively. Input pattern element that represents the situation, a ₁ (a ₂ )
Is the floor difference between the floor where the new hall call is generated and the current floor of Unit 1 (2), and b ₁ (b ₂ ) is between the current floor of Unit 1 (2) and the floor where the new hall call is generated. Number of calls in charge, c
₁ (c ₂ ) represents the number of hall calls assigned to Unit 1 (2) beyond the floor where the new hall call is generated, and d ₁ (d ₂ ) represents the current number of passengers in Unit 1 (2) .

It should be noted that all of these input pattern elements are calculated by the input pattern calculation means shown in FIG. Sixth
The neural net 8 in the figure is not shown in the figure,
An input layer composed of units corresponding to the respective input pattern elements a _{1 to} d ₁ and a _{2 to} d ₂ , an intermediate layer composed of an appropriate number of units (not limited to one layer), and outputs A ₁ and A It consists of an output layer consisting of units corresponding to ₂ . Where A ₁
Is an output signal indicating the allocation suitability of the No. 1 machine, A ₂ is an output signal indicating the allocation suitability of the No. 2 machine, and these A ₁ and A ₂ are inputted to the allocation judging means 9 of FIG. 5 and the allocation signal 10 is outputted. It

In the above configuration, the characteristics of the neural net must first be determined by setting the weights between the units in order to actually perform the allocation. This weight is set using learning by back propagation known as a general setting method. This learning is performed as follows.

First, the weights between the units are randomly set to appropriate values. On the other hand, many learning samples are created from actual driving examples. This learning sample assumes, for example, one combination of a new hall call and the states of Units 1 and 2 at that time, and in this state, the number of units to allocate is determined by the expert's judgment or using simulation or the like. If the No. 1 machine is optimal, the output target of the output A ₁ corresponding to the No. 1 machine is set to 1 and the output target of the output A ₂ corresponding to the No. 2 machine is set to 0 for each input pattern element at that time. It was created as. Then, using the first learning sample, the input pattern elements a _{1 to} d ₁ and a _{2 to} d ₂ when a new hall call is generated are calculated from the states of Units ₁ and ₂ and each unit of the input layer of the neural network is calculated. To enter. This input data is processed sequentially from the input layer to the output layer,
Since each unit has the input / output characteristics shown in FIG. 3, as a result, the outputs A ₁ and A ₂ are always values of 0 to _1, but as described above, the weights are initially set randomly. , The values of the outputs A ₁ and A ₂ are different from the output target. Therefore, the difference between this output target and the actual output is taken as an error, and the weight is corrected using this error from the output layer to the input layer. Since the calculation of the correction of the weight is well known, its explanation is omitted here.

When the correction of the weight between each unit is completed in this way, each input pattern element is then calculated and input by the second learning sample, and the error between the output target at that time and the actual output is again calculated between each unit. Correct the weight of. When the same procedure as described above is repeated using a large number of learning samples in this way, the above error becomes sufficiently small and eventually converges. When the learning is completed, each weight converges to its own value, so if it is fixed, the weights between the units of the neural net used for actual driving are set respectively.

After that, this is used for the actual allocation. That is, when a new hall call is generated, the input pattern elements a _{1 to} d ₁ and a _{2 to} d ₂ are calculated by the input pattern calculation means 7 from the conditions of the first and second units at that time, and are input to the neural network 8. To be done. In the neural net 8, each input pattern element is sequentially processed from the input layer to the output layer, and the outputs A ₁ and A ₂ are input to the allocation determining means 9. The assigning means 9 compares the outputs A ₁ and A ₂ and outputs the assigning signal 10 to the machine whose value is closer to 1. At this time, each weight of the neural network 8 is a learning sample. Since the values are learned and used to be set to converged values, the allocation is performed according to the same criterion as the learning sample.

In this way, by using a neural net, it is possible to automatically create a judgment system that determines the optimal assigned car according to various traffic situations simply by creating many learning samples and repeating learning. Therefore, it is possible to perform extremely advanced allocation control.

[Problems to be Solved by the Invention]

By the way, the efficiency of the learning and the accuracy of the assignment after the completion of the learning depend largely on what is selected as each input pattern element to the neural network, that is, as the parameter indicating the situation of the entire group when a new call occurs. It depends.

Theoretically, this parameter may be anything related to the allocation decision, such as the operating status of each elevator, the status of calls, the status of waiting passengers at each hall, etc. It is expected that the more closely related the parameters are to the allocation decision, the higher the accuracy of the allocation.

Therefore, as parameters closely related to the determination of allocation, it is considered to use parameters such as the estimated waiting time of each hall call and the expected number of waiting passengers at each hall, which are often used in the conventional allocation method using the evaluation function. However, using a parameter that is such a predicted value has the following disadvantages.

That is, such a parameter is not a direct parameter representing a fixed value that can be directly measured, such as a car position or an elapsed time after a call is generated, but is calculated based on a direct parameter. It is an indirect parameter that represents the predicted value and may deviate significantly from the actual value. Therefore, when these indirect parameters are used as input pattern elements to the neural net, the learning convergence slows down. However, there was a problem that the learning effect deteriorated.

[Means for Solving the Problems]

The present invention solves the above problem, that is, by using an indirect parameter as an input pattern element to a neural network, highly accurate assignment is made possible, and the learning efficiency is not deteriorated by this. It was made to do, and the feature is that
A second neural net is provided separately from the assignment neural net, and a direct parameter representing a directly measurable value is input as an input pattern element to the second neural net to represent a predicted value. The point is that an indirect parameter is obtained as its output, and the indirect parameter and the direct parameter are input to the allocation neural net as input pattern elements.

〔Example〕

 An embodiment of the present invention will be described below with reference to FIG.

FIG. 1 is a diagram corresponding to FIG. 6 showing the relationship between the neural net and the input pattern elements. The same components as those in FIG. 6 are designated by the same reference numerals.

In FIG. 1, the input pattern elements a _{1 to} d ₁ and a _{2 to} d ₂ are direct parameters such as the floor difference between the floor of the new call and the current floor of each unit, as described above. , They are input to the neural network 8a for allocation and the second
It is also input to Neural Net 8b. In the second neural net 8b, these direct parameters are input, and the indirect parameters S ₁ , S ₂ and t ₁ ,
get t ₂ . This parameter S ₁ (S ₂ ) represents the estimated waiting time for a new landing call for Unit 1 (2), and t ₁ (t ₂ ) is the estimated number of passengers when Unit 1 (2) services a new call. Are input as input pattern elements to the neural net 8a, and are also used for display on the hall.

In the above-described configuration, the output of the neural net 8b is necessary for the operation of the neural net 8a, so that the neural net 8b is first learned. In order to carry out this learning, a large number of learning samples are created from actual driving examples, as in the case described above. In this case, the output S ₁ , S ₂ and t
₁ and t ₂ are predicted values, but in the actual driving example, the waiting time and the number of passengers can be obtained as fixed values when the call is serviced, so a new call is generated as a learning sample. Pattern elements a _{1 to} d ₁ and a _{2 at the time}
A combination of ~ d ₂ and the actual waiting time and the number of passengers for it is used as the output target.

When the learning of the neural net 8b is completed, the learning of the neural net 8a is performed next. This learning is the same as that described with reference to FIG. 6, but a _{1 to} d ₁ and a ₂ which are direct parameters as the input pattern elements.
Not only ~ d ₂ but also indirect parameters S ₁ and S ₂ and t ₁ and t _{2 which} are closely related to the allocation decision are input, and these indirect parameters are transmitted by the neural network 8b. Learning is performed efficiently because the value is highly reliable.

When the learning of the neural nets 8a and 8b is completed in this manner, it becomes possible to allocate the actual call. That is, when a new hall call occurs, the input pattern element a _{1 at} that time
~ D ₁ and a ₂ ~ d ₂ are input to the neural net 8b,
S ₁ , S ₂ and t ₁ , t ₂ are output. And this S ₁ , S ₂ and t
₁ and t ₂ are input to the neural net 8a together with a _{1 to} d ₁ and a _{2 to} d ₂ as input pattern elements, and if the hall call is assigned to either car by its output A ₁ and A _2. It is decided whether it is good or not, and the car is assigned.

In the above embodiment, the learning operation of the neural network is only performed in advance before the operation of the elevator system, but by continuing the learning operation even after the operation of the elevator system, the traffic situation unique to each building Can be flexibly dealt with.

In the above embodiment, the neural net is used to make the allocation decision. However, the neural net is used only to obtain indirect parameters, and the conventional evaluation function or expert system is used to make the allocation decision. Even when using, it is possible to expect an improvement in allocation accuracy.

〔The invention's effect〕

According to the present invention, the indirect parameter representing the predicted value closely related to the determination of the allocation is obtained by the second neural net provided separately from the neural net for allocation. Can be used for efficient learning and can be used as a highly reliable parameter.

Moreover, by using this highly reliable indirect parameter as the input pattern element of the neural network for allocation, the learning efficiency of the neural net for allocation can be improved, and highly accurate allocation can be performed. it can.

Further, the highly reliable indirect parameter obtained by the second neural net can be used for other than the allocation determination, for example, for displaying waiting time at the hall.

[Brief description of drawings]

FIG. 1 is a diagram showing the relationship between a neural net according to the present invention and input pattern elements, FIG. 2 is a diagram showing the structure of a unit that models a neuron, and FIGS. 3 (a) and 3 (b) are the input units. FIG. 4 is a diagram showing an example of the output characteristics, FIG. 4 is a diagram showing the structure of a general neural net, and FIG. 5 is a diagram showing an embodiment of the overall configuration when the neural net is used for allocation control. FIG. 6 is a diagram showing the relationship between a conventional neural net and input pattern elements. 8a ... Assignment neural net 8b ... Second neural net a _{1 to} d ₁ ...... Input pattern elements a _{2 to} d ₂ representing direct parameters corresponding to Unit 1 Corresponding to Unit ₂ Input pattern element representing a direct parameter S ₁ , t ₁ ...... Input pattern element representing an indirect parameter corresponding to Unit 1 S ₂ , t ₂ ...... Input pattern representing an indirect parameter corresponding to Unit 2 element

Claims (1)

[Claims] 1. A plurality of elevators are put into service for a plurality of floors, and various parameters representing the situation at that time are assigned as input pattern elements to a neural net for allocation for a newly generated hall call. In an elevator group supervisory control device for inputting and selecting an optimum car according to its output, a second neural net is provided in addition to the above-mentioned neural net for allocation, and the second neural net is provided. A direct parameter representing a directly measurable value is input as an input element to the net, an indirect parameter representing a predicted value is obtained as its output, and the indirect parameter and the direct parameter are input patterns. An elevator group supervisory control device, characterized in that the elements are input to the allocation neural net.