FR2968110A1 - Method for managing traffic on road, involves identifying equivalent traffic situation based on forecasted traffic demand, and publishing forecast of state of traffic on short term basis based on traffic data of equivalent traffic situation - Google Patents

Abstract

The method involves preparing (E2) a given set of traffic data measured in traffic situations, and storing the prepared data in a historical database. A traffic demand is forecasted (E3) based on the data stored in the historical database, and an equivalent traffic situation is identified (E4) based on the forecasted traffic demand. The forecast of a state of traffic is published (E5) on a short term basis based on the set of traffic data under the equivalent traffic situation. An independent claim is also included for a system for management of road traffic, comprising a database.

Description

The present invention relates to a traffic management method for a road network. It also relates to a management system implementing the method. In the state of the art, traffic management systems of a road network are known in which road traffic condition sensors are arranged in relation to the roads of the managed network. These sensors provide information about various traffic state parameters lo and are then processed by a central processor to generate traffic management information based on predetermined rules. In particular, road traffic management systems are known which generate information for users and which are displayed on roadside billboards or on gantries over the road. These billboards make it possible to inform the driver, who passes near the visual panel, to have an estimation of the duration of the journey that awaits him to a specified destination, such as the passage to an exit of the 20 road that he borrows. In other applications, information concerns, for example, the average speed of traffic on a road axis, the cumulative length of road congestion on the network or other information which is transmitted over a communication network for users equipped with terminals, such as a computer connected to the Internet or a mobile phone connected to a digital telephone data network. This information is intended to encourage users of the road network to modulate their use of the road network according to the projected traffic. In other applications, the traffic forecast information is intended to control means for closing or temporarily opening access routes to a determined road axis according to the forecast traffic data. Thus, when a road axis reaches a certain degree of congestion, it is known to control a traffic light to stop the entry of additional vehicles on the axis during congestion. Such a control of a fire requires that the very short-term forecast of traffic variation has been as accurate as possible. However, there exists in the state of the art a need for a more efficient road traffic management method.

It is an object of the present invention to provide a method for managing road traffic based on a method for predicting the evolution of traffic in the very short term. To this end, the present invention relates to a method for managing road traffic characterized in that it comprises the steps of: - Preparing a determined set of measured traffic data and storing the measured data in a history of traffic situations at save in a history database; - Forecast traffic demand on the historical database; - Identification of an equivalent situation in the history of past traffic situations based on the expected traffic demand; and - Editing a forecast of a very short-term road traffic condition based on the equivalent traffic situation in the form of a set of expected traffic data. According to other characteristics of the method: the step of identifying an equivalent traffic situation in a history of past traffic requests consists in: determining a traffic request identification period; - determining at least one equivalent situation in a situation history recorded in the situation history database, on the traffic request identification period in the history of the observed requests; the step of predicting the traffic request comprises a step of determining a nominal traffic demand profile for each type of day established on a predetermined calendar typology, the step of determining a demand profile comprising a determination step: - of a set of flow averages over the entire history; - a set of weighted flows for the days of a given calendar typology, for each day of this typology on the basis of historical data according to a relation of the form: <qh> = average (q, h) ; j of calendar typology c and hp <h <hF where h is a time stamp in the time slot defined in each day type day c and mean () is a mean calculation function; a median of the weighted flows for each type of day of the typology determined according to a relation of the form: pc, h = median (p, h); for each day j belonging to the class c in the calendar typology Ecalendaire, and hp <h <hF, taken as nominal demand profile for the type c of day; the method comprises a step of selecting the expected nominal demand recalculated by searching for the smallest distance between the observed flow rates Q and respectively the expected nominal demand D, the expected nominal demand recalibrated by an average D, and the expected nominal demand recalibrated by a median D2, and as a function of distance, a cumulative distance function is used, as sum of the partial distances each calculated as the distance between two elements of the same rank, according to a predetermined norm; the step of identifying an equivalent situation in the history of past traffic situations on the basis of the expected nominal traffic demand comprises a step of: pre-sampling of the equivalent situations; then - sampling of equivalent situations based on pre-sampling; - the pre-sampling stage of equivalent situations consists in finding the smallest distance between candidate situations and the situation observed during the last observation which precedes the very short-term forecast period, on the basis of the cumulative distance method as the sum of the partial distances each calculated as the distance between two elements of the same rank, according to a predetermined norm, and applied to n situations Si having the same time slot in the day as that of the situation observed during the last observation preceding the very short-term forecast period, taken on at least one of three quantities of flow, occupancy rate and / or average speed, and nominal traffic demand observed in situations close to So, and previously subjected to statistical centering and reduction processing; - the step of sampling equivalent situations consists of looking for the smallest distances between candidate situations taken from the situations obtained by pre-sampling and the situation observed during the last observation which precedes the very short-term forecast period, on the basis of the method of measurement distances taken from at least one of three quantities of flow, occupancy and / or average speed, and nominal traffic demand observed in situations close to the situation, and the step of editing a forecast of a very short-term road traffic condition based on the equivalent traffic situation in the form of a set of planned traffic data includes a step of constructing the forecast in the very short term from the traffic measurements of the day I of the history which presents the situation having the smallest distance by measurement among the situation s identified at the sampling stage of equivalent situations; the method further comprises a step for executing prediction actions which comprise: the dissemination of road forecasting information; and / or - the control of traffic control actuators as traffic lights; on the basis of the edition of a forecast of a state of very short-term road traffic. The present invention also relates to a road traffic management system for implementing the method according to the invention characterized in that it comprises: a database of past traffic measurements; a set of traffic data sensors to be measured; means for processing traffic measurement data; a means for calculating a traffic request at a given time stamp; a means of identifying an equivalent traffic situation; a means for calculating a forecast of a very short-term road traffic state; and at least one device for exploiting very short term traffic condition forecast data. According to other features of the system: - it includes a traffic management module includes means for editing traffic forecast information bulletins to terminals; the terminals are: mobile phones or similar devices, and the communication network comprising a data network over a mobile telephone network; and / or onboard terminals on vehicles connected to the module by a mobile telephone network, the onboard terminal being adapted to such a mobile telephone network as a radio-frequency receiver for a vehicle and the communications network comprising a network of mobile telephones. voice or digital traffic information such as the RDS-TMC network; - The traffic management module is connected to a network of traffic information displays, and then includes a road information editor adapted to communicate with the displays; the traffic management module is connected to a network of traffic actuators comprising, in particular, traffic lights, vehicle speed limitation panels, access barriers to one or more road network lanes, the traffic control module; traffic management comprising means for generating control signals for controlling the traffic actuators; the traffic management module is connected by means of communication to toll barriers, so that the charging of the access or the exit of a vehicle depends in particular on the forecast of very short-term traffic generated by the traffic management system of the invention, and the module comprises means for determining a pricing control message intended for payment machines and / or for displaying on displays reserved for this purpose at each tollgate and provided for for the users and / or the personnel charged with the collection of the tolls.

Other features and advantages of the present invention will be better understood from the following description and the appended drawings in which: FIG. 1 represents a flowchart of the method of the invention; FIG. 2 represents a block diagram of one embodiment of a traffic management system according to the invention. In Figure 1, there is shown a flowchart of the method of the invention. The process is based on a very short-term forecasting method. In one application, the forecast duration or forecast horizon is 4 hours from the last measurement observed. In addition, it is noted that the method of the invention is based on the knowledge of a history of traffic situations recorded in a database of traffic situations, and on the measurement of the traffic situation before executing a traffic situation. very short term forecast.

In the invention, a timestamp is a data structure, consisting of a date of a day of the year and a time of the day, which thus makes it possible to consider identical time slots for each day. The situations are thus compared or evaluated on the basis of the same time slot for each calendar day. Thus, the traffic status at time "08:00" will be compared on calendar days having a comparable traffic characteristic. A traffic situation is characterized by a set of measurements obtained by observing road traffic in at least one measurement point on the road network. In one embodiment of the method of the invention, the assembly comprises three measurements which are: the flow rate, which informs the distribution of the vehicles over time; - the occupancy rate, which informs of the proportion of time during which a measurement point is occupied by a group of vehicles of at least one vehicle; and - the average speed, which informs the average time of the instantaneous speeds of the vehicles. For the term "measurement", the term "magnitudes" is used to designate and distinguish the types of measurements found among flow, occupancy and average speed. In practice, a measurement collected at a given time stamp includes a value for one, two or three of the quantities mentioned (depending on the type of network: generally in an urban network, only the rate Q and the occupancy rate To are recorded, while in the interurban network, the rate Q, the occupancy rate To and the average speed V are recorded). In a particular embodiment, the inventors were able to determine that a good compromise between the acquisition cost and the quality of the forecast was achieved by the following definitions.

The traffic flow rate is defined by the relation: q = n / (tb - ta) in which: q is the flow rate at the curvilinear abscissa x; n is the count of the number of vehicles passed at the x-axis measurement point x between the timestamps ta and tb, arbitrarily defined by tb> ta. The occupancy rate i is defined at the same point of measurement of curvilinear abscissa x from the same time stamp of the beginning of measurement by the relation: i = L * p in which: L is a predetermined constant which represents the average space occupied by a vehicle on the network; p is the concentration of the vehicles at the x-axis measurement point x, between the time stamps ta and tb of measurement. The mean velocity v is defined as an arithmetic mean of the vehicles that have passed the point of measurement of curvilinear abscissa x by the relation: v = sum (v;; i IN [1, n]) / n; Where n is the number of vehicles passed between the two instants ta and tb at the x-axis measurement point; and V; is the instantaneous velocity taken from the vehicle Lerne passed at the x-axis measurement point x between the time-stamps ta and tb.

In the traffic management system implementing the method of the invention, the measurement sensors of the three quantities of flow, occupancy rate and average speed are formed according to the available road infrastructure. Thus, the traffic management system may comprise: single or double inductive loops, embedded beneath the roadway and connected to inductive signal processors to deduce the two or three quantities mentioned above, depending on the number of loops deployed; piezoelectric sensors, implanted under the roadway and connected to piezoelectric signal processors to deduce at least one of the three quantities mentioned above; - video cameras working in the visible spectrum and / or infrared for night vision, implanted or on one of the edges of the track, or on a gantry above the roadway, and connected to video signal processors to deduce the three quantities mentioned above; sensors associated with the toll boxes, in particular for counting the vehicles and deriving average speeds on routes between two boxes, these data being often used to test assumptions of validity of reconstruction of missing traffic data; - sensors associated with databases held by mobile operators and / or geolocation. In the latter case, it is indeed possible to reconstruct partial data including the counting of vehicles equipped with or carrying mobile phones, which can be located in cell phone cells by the mobile operator. Similarly, geolocation terminals cooperating on the computer machines of a geolocation operator, instant measurement databases and historical data are also available from these geolocation operators. They make it possible by known means to reconstitute a traffic flow rate, vehicle counts per unit of time, average speed data and therefore, by deduction, an estimation of the flow rate and / or occupancy rate data. measuring points which are determined solely by calculation instead of, as is the case for the aforementioned physical sensors such as inductive loops, to be at the curvilinear abscissa of the position of the inductive loop on the track. In Figure 1, there is shown a flowchart of the traffic management method of the present invention. The method essentially comprises four main steps which are: a step E2 of storage of the raw data measured by the traffic sensors in a history of traffic situations to be recorded in a database of history and preparation of these data (described by the following) ; a step E3 for predicting the traffic demand on the basis of the measured traffic data and a calendar typology; a step E4 of identifying an equivalent situation in the history of past traffic situations on the basis of the expected traffic demand and the situation under observation; and a step E5 of editing a forecast of a very short-term road traffic condition on the basis of the equivalent traffic situation in the form of a set of expected traffic data. The database provided in step E2 contains a history of traffic situations. A traffic situation contains a list of measured values at each measurement point on the network managed by the method of the invention. Referring to Figure 2, there is shown a portion of a road network containing an axis 1 (or roadway) on which flows a vehicle stream. The vehicles in the management method of the invention are not taken into account at the microscopic level, but the measurement quantities of the flow thus constituted are observed. The road network contains, of course, several axes such as axis 1, each having a single direction or two directions of traffic, then designated respectively in ascending and descending direction of traffic; there is also at least one lane per direction of traffic existing on an axis. Neither was there a road intersection, interchange, access or exit lane. But a measuring point can be placed anywhere on a road network. A set 2 of traffic data sensors is disposed along each path as the axis 1 of the road network. According to the principle of the invention, the sensors 2 necessary for the management of traffic are determined. These sensors are placed at specific measuring points. The position of the measuring point is indicated by a curvilinear abscissa measured from a reference point of the road network (not shown). The number and type of quantities measured and therefore the type of sensor provided at each measurement point depends, as indicated above, on the type of network, such as an urban or interurban network. In FIG. 2, three measurement points 21, 22, 23 have been represented and their position is respectively designated by the curvilinear abscissae x ,, x2, x3. Of course, a higher number of measurement points 2 can be used. The increase in the number of measurement points is made necessary on a large road network with a large mesh with a large number of intersections or different lanes. In the analysis made by the inventors, the measurement quantities of the road traffic are at each measurement point: the number of vehicles passing between two timestamps ta and tb; - the instantaneous speed of the vehicles at the point of measurement; and - The concentration of vehicles. However, the analysis led the inventors to use three quantities of observation derived from the quantities of measurement with which they will be confused in the rest of the text and which are, respectively: the observed flow rate; - the average speed observed; and the occupancy rate observed according to the calculations which have been described above.

When all the observation quantities at a time stamp have been collected on all the measuring points of the road network, a traffic situation is available to be recorded in a database of historical records. traffic situations, after a treatment that will be described later. In the analysis made by the inventors, another notion is of great importance: that of confrontation between a demand for traffic and a supply of traffic. According to the invention, the traffic demand is a quantity based on an average of vehicle flows. The nominal traffic demand is constituted on the basis of all vehicles traveling on the road network at the observation time table ta. It depends entirely on the statistics of road users' behavior, their number and their itinerary intentions. The nominal demand is then the most frequently measured demand, with fixed context. With respect to the traffic demand, the traffic supply consists of the state of the network, a state that is directly reflected in the observation variables. The principle of the very short-term forecasting method consists, according to the invention, in comparing an estimate of the traffic demand with the history of the traffic situations, representing over time the evolution of the three observed rate observed occupation and observed speed. When the very short-term forecast has been executed over a very short-term forecast horizon, it is possible to extend the history of the situations observed by a traffic forecast represented by the list of values of the quantities observable over the period from the last traffic watch timestamp at the next timestamp, separated from Th the very short-term forecast horizon, ie ta + Th.

It should be noted that in the method of the invention, time period selections are made in the history of the traffic situations for which the traffic variables calculated according to the invention are taken as explanatory variables of the traffic. The supply or demand of traffic is essentially a flow of vehicles on the road network at a given point in the network. One of the aspects of the invention has been to identify an explanatory period during which the observed variables really have an explanatory value. This concept of explanatory period is similar to the forecast horizon, which will also be used, which is the validity period of a forecast of traffic situations.

In one embodiment, the explanatory period Te is of a duration of 8 hours, while the forecast horizon Th is 4 hours. The principle of the method of the invention then consists of looking for a past traffic situation, equivalent to the traffic situation observed. Then, on the basis of this past equivalent traffic situation, a very short-term forecast is derived from the evolution of past traffic situations which follow the observed equivalent situation over the period constituting the very short-term forecast horizon. Th.

Returning to FIG. 1, step E2 performs a preparation processing of the measured data. Several steps are planned separately or in combination, and which are: - a step of disqualification of outliers; - a step of disqualification of the beginnings and ends of outliers; a step of smoothing the measurements; and a step of reconstituting the absence of measurement. The disqualification of outliers consists of performing on each traffic measurement a test of a disqualification criterion.

Among the disqualification criteria, the method of the invention, in a preferred embodiment, retains: the disqualification of negative or zero flow rates; - the disqualification of strictly negative occupancy rates; and - the disqualification of the negative average speeds, zero or higher than a threshold of disability of the measured speeds. In one embodiment, the verification of only one of the three criteria makes it possible to disqualify the complete measurement. The disqualification of the beginning and end of outliers disqualifies the last previous measure and the first measure following a disqualified measure during the disqualification stage of the outliers. Thus, if the measure A (t ;, xi) is declared aberrant during the disqualification step of the outliers, then it is decided to disqualify in the next step of disqualification of the beginnings and ends of outliers the previous measure. A (ti_1, xi) and the next measurement A (t; +1, xi), for the measuring point 21 (Figure 2) disposed at the abscissa xi. The measurement considered of course relates to one of the three quantities observed according to the method of the invention: flow rate, occupancy rate or average speed. Thus, a disqualified measure on at least one of these three magnitudes invalidates each of the values recorded on the three magnitudes at a given time stamp. In one embodiment, the method of the invention during the preparation of the data of step E2 (FIG. 1) also includes a step of filtering the measurements so as to eliminate the measurement noise while preserving the traffic characteristics. Thus, the filtering can be adjusted to ensure that after the data is prepared, the traffic situations recorded in the traffic history database 10 (FIG. 2) make it possible to account for the characteristics of the traffic. Preferably, the filtering performed is a smoothing by calculation of a mean value. If then we have a series of successive measurements in time, at timestamps t; _p, ti_p + 1, ---, t ;, we calculate the arithmetic mean of the measurements A (t ;, xi) at the point d the abscissa xi on the road network on the p + 1 previous values, so that the smoothed (or filtered) value is B (t ;, xi) defined by the relation: B (ti, xi) = Sum (A (t ; k, xi); i = [O..p]) / (p + 1) This operation is referred to as inherited smoothing to oppose it to the symmetrical case not retained in the embodiment described here, where the Smoothed value is taken at the beginning of the smoothing period. In one embodiment, the method of the invention during the preparation of the data of step E2 (FIG. 1) also includes a step of reconstituting the missing measurements. A measurement may be missing because of the failure of a measuring point or one of its sensors, a communication network fault that links the measuring points to a central computer, or a disqualification by one previous steps. In such a situation, a linear interpolation is made between the remaining measurements that frame the missing measurement.

Such an interpolated measure Mi corresponds to a measurement missing from the timestamp ti and the two remaining measures are Ma at the time stamp, last measure before a period of missing data, and Mb at the time stamp tb, the first measure after a period missing measures. In one embodiment, it is decided to apply a further criterion of maximum permissible interpolation period Ainterpol which is the maximum duration during which reconstitution of missing data is possible. Such a threshold value must be initialized in the method of the invention and it is set to a predetermined value which, in particular, respects the capacity of the method to carry out a very short-term forecast using interpolated data. For example, the interpolation limit period is set to 18 minutes, in a very short-term forecast over a period of four hours and for data collected over a period of six minutes. Typically, we will take an interpolation limit period worth three times the collection time of a measurement. The interpolation condition is then written by the relation 0 <tb - ta <Ainterpol and the classical interpolation relation is then: for all ti 1 ta <ti <tb Mi = Ma + (ti - ta) * (Mb - Ma) / (tb - ta) for any magnitude of the measurement observed M in the context of the traffic management method of the invention, such as the flow rate, the occupancy rate and the average speed, at each measurement point d curvilinear abscissa x; of timestamp ti between two correct observing timestamps ta and tb, respecting the interpolation condition.

During a last measurement time stamp, we have all the information necessary to describe a traffic state necessary to accurately guide the trend of traffic evolution in the very short term. For this purpose, the present invention comprises a step E3 for predicting the traffic demand on the historical data base. The history contains all the traffic reports or situations that are successively recorded.

Step E3 begins with a step of determining a type of day according to a predetermined calendar typology. In one embodiment, the typology criterion consists in grouping the days according to the criteria of social activity of the users by distinguishing the days off, the days of work, the holidays, etc. It is possible to set freely during the initialization of the process categories of days of the year that constitute a calendar typology. The calendar typology is designated by Ecalendaire, so that, especially in the exploration of the history database of traffic situations, each date can be assigned to a type of day, that is to say that the type or category of the calendar on which the day is considered can be determined. A traffic request profile is then established for each type of day of the selected calendar typology Ecalendaire- The traffic request profile is established as follows. The averages <qh> of the flows are computed over the entire history for time stamps h between hp and hF, time stamps at the start of the time slot and at the end of the time slot, arbitrarily fixed in each day. Preferably, hp corresponding to the time "00:00" and hF to the hour "24:00" of the day j of day type c. For all time stamps h recorded for a day d of day type c in the history database, calculate: <qh> = average (q; h); j of day type c and hp <h <hF Each <qh> represents an average hourly flow recorded in the history of the measuring point. For example, q (02h00) = average (gi (02h00)), i = [O..n], where n is the number of days in the history. We then perform the calculation of the weighted flows for each type c of the Ecalendaire calendar typology, according to a relation of the form: p;, h = q;, h / <qh>; for day type day c and hp <h <hF Finally, we calculate the medians of the set of weighted flows, for each type of day c of the calendar typology, according to a relation of the form: pc, h = median (p, h); for each day j belonging to the class c in the calendar typology Ecalendaire, and hD <h <hF It is noted that the flows of all flows used for the calculation of the medians are calculated independently for each measurement point. They do not depend on the abscissa. This gives the nominal demand profile for each type of day of the Ecalendaire calendar typology in the daily time slot. The nominal demand profile for the type c calendar day is a set of data {pc, h; h hd. <h <hf}. There are as many nominal demand profiles as there are types c in the calendar typology Ecalendaire- The median function () used to define the medians of the weighted flows allows to select the term pc, h, which divides into two halves equal to the set of values {(p;, h); hD <h <hF} after it has been ordered in increasing or decreasing values. In the method of the invention, for the step E3 for predicting the traffic demand on the history data base, a step of identifying the optimal nominal demand is then carried out. From the previous step, we have the list of nominal demand profiles for each type of day c of the calendar typology. The next step, in one embodiment of the method of the invention, is, for a day observed jo, to determine an optimal nominal demand profile. To obtain such an optimal demand profile, the invention consists in calculating the weighted flows for any time stamp h between hD and ho, respectively the time stamp of the start of day time slot, and the time stamp of the last observed time slot, according to the relation: p; o, h = q; o, h / <qh> for the day of the last observation The set of weighted flows {p; o, h; h IN [hD, ho] is the flow profile of the observed day jo. The optimal nominal demand profile is then identified by comparing the set of nominal demand profiles established for each type of day, which is noted: {{pc, h, h hD- <h <hF}, c in Ecalendary} to the debit profile of the observed day jo.

The step E3 of identifying a traffic request then comprises a step of forecasting the actual nominal request. This operation consists in establishing a nominal demand forecast from the optimal nominal demand profile obtained in the previous step, in step E3. The expected traffic demand is called a dh list, valid for a period including the timestamps counted from the forecast start date, that is to say, when operating the method the horodate ho of the last observed situation. and loaded into the history of the database 10, and the very short-term forecast period having a duration or forecast horizon Ts of four hours in a preferred embodiment. In one embodiment, the current term dh of the expected demand d is given by the relation: dh-pi, h * <qh> where: h is the current timestamp of the calculated term in the very short-term forecast period Ts; pi, h is the h-th term of the optimal nominal demand profile selected in the previous step; and <qh> is the average of the flows on a predefined time slot, for the timestamps h between hp and hF, of start and end of time slot. In one embodiment, the expected traffic request is then processed to fit the history data. Preferably, the processing is a recalibration operation of the nominal demand forecast as a function of the flow rate observed over the explanatory period. The resetting operation is used to scale the request. According to the invention, two types of registration may be selected: by means; - by medians. In this step of resetting the planned demand, the following successive operations are executed: calculation of the differences of the flow rates observed two by two; - calculation of the deviations of expected requests two by two; - calculation of the ratio of the deviations of the requests to the deviations of the flows; Fmoyenne - calculating the divergence of the ratio of deviations by mean and median Fmédiane, respectively by calculating the averages of the ratios of the gaps and the medians of the reports of the deviations; calculation allowing the scaling of the expected demand according to the divergences obtained previously, ie dh, 1 for the mean F average and dh, 2 for the median Fmedian, according to a relation of the form dh, 1 = dh / F average and dh, 2 = dh / Fmédiane, for all time stamps h, in the current time slot before the horodate ho of the last traffic observation before the forecast period; calculating an offset p between the observed flow rates and the scaled demand in each of the p1 or median p2 cropping cases, according to the respective relations: p1 = sum (qh - dh, 1; hp < h <ho) and p2 = sum (gh - dh, 2; hp <h <ho); calculation of the adjustment of the volume of nominal demand forecast and recalculated according to the two mean and median methods, according to the respective relations Dh, 1 = dh, 1 + p1 and Dh, 2 = dh, 2 + p2, for the horodates h between hp <h <ho; selecting the expected nominal demand adjusted by searching for the smallest distance between the observed flow rates Q and the expected nominal demand D, the expected nominal demand adjusted by the average D1 and the expected nominal demand adjusted by the median D2. The expected nominal demand is therefore one of the three nominal requests D, D1, and D2 resulting from the registration step. As a function of distance, any suitable standard can be used. In one embodiment, the traffic management method uses a cumulative distance function. A cumulative distance function is the sum of the partial distances, the terms of which, in one embodiment, are weighted by coefficients determined by calibration on a test history. A partial distance is calculated as the distance between two elements of the same rank in two sets of data whose cumulative distance is to be calculated. In the case of the selection selection of the expected nominal demand, the three sets of data of the same size are the expected nominal demand before registration D, the expected nominal demand recaled by the average D1 and the expected nominal demand recaled by the median D2. , which are compared by the cumulative distance to the data set of the flow rates Q. Each of the terms of the three sets of data is described by dh, d ,, h, d2, h, over the period [hp..ho] of the last time slot.

In the following road traffic management method according to the invention, the identification step E4 of an equivalent situation in the history of past traffic situations on the basis of the expected traffic demand comprises a step of: - pre-sampling of equivalent situations; - sampling of equivalent situations on the basis of pre-sampling. The identification step E4 of an equivalent situation in the history of traffic situations, actually observed and recorded in the historical database 10 (FIG. 2), has the advantage of a very high speed of detection. identification and to obtain a very small divergence between the two close situations. The pre-sampling step, which is the first step comprising the identification step E4 of an equivalent situation, comprises a step of selecting a sample of candidate situations for equivalence, then a step of selecting the sample of the best equivalences. The step of selecting a sample of candidate situations for equivalence consists in finding the smallest distance between candidate situations and the situation observed during the last observation preceding the very short-term forecast period. Note that we use the same explanatory period, that is to say for all time slots h days retained, such as: ho - Te 5 h 5 ho, for the three quantities of flow Q, occupancy rate To and average speed V, and such that ho 5 h 5 ho + Th, for the request D.

Preferably, the aforementioned cumulative distance method is used as the sum of the partial distances each calculated as the distance between two elements of the same rank, according to a predetermined standard. We consider two traffic situations taken as data sets of the same size, respectively So for the last situation observed at the horodate ho, and the previous observed situations in the history with the number of n, ie Si. Each situation is defined, as has been described, for three observed traffic quantities of flow, occupancy rate and average speed. The n situations If retained in the analysis are those which have the same time slot in the day as that of So. In one embodiment, the three observed magnitudes of the situations approximated to the So situation are subjected to a conventional statistical centering and reduction treatment, which makes the data independent of the unit and the scale, and finally to give the quantities treated the same average and a similar dispersion, so that the quantities become comparable. These operations therefore require the calculation of an average and a standard deviation for each observed quantity of flow, occupancy and average speed. The cumulative distance is then performed on the three observed centric-reduced rates of flow, occupancy and average speed and on the projected centric-reduced demand. Now, if situations are taken on the same time slot that is of a duration Te while the expected centric-reduced demand is taken on the forecast horizon Th which is different. The cumulative distances are returned to the same time dimension using the fact that the observed magnitudes are measured for a predetermined time z; so that the cumulative distances of each of the situations Si to the situation So are averaged over the same number of data. By designating by M one of the three quantities observed in the Si or So situation, and by N the expected demand, and denoting by M 'the corresponding centered-reduced value or N' the planned centric-reduced demand, one can write cumulative distances taken into account by: dM; = (z / Te) * Sum (lMh,; - M'h,; l; ho - Te <h <ho) dN; = (z / Th) * Sum (INh , - Ho, he; ho <h <ho + Th) The step of selecting the sample of the best equivalences is then performed to select among the distances dM; and dN; calculated in the previous step the smallest distances. We then constitute an ordered set D of distances dM; and dN; sorted in ascending order. Then, we select the first Np elements of D, pre-sampling set observed situations equivalent to the situation So, with Np a predetermined number selected in step E, initialization of the process.

We thus generate a set Rp of the references of Np situations equivalent to So. We then perform a final sampling step of the equivalent situations among the references Rp Np situations equivalent to So last situation observed and this set being obtained in the previous step cumulative distances. During this step, we look for the situation offering the best equivalence with respect to the last observed situation So using a distance function by measurement, distance of better precision than the distance function used for the calculation of the together Rp of the Np references of the best situations equivalent to So. In a particular embodiment, a distance function is used by measurements and the following steps are performed for this purpose: - calculation of distances by measurements; - selection of the best equivalence.

The distance by measurements used in the final sampling stage of the equivalent situations is limited to the set Rp of the Np references of the best situations equivalent to So, which is smaller than the set D of the situations of the same time slot for which the cumulative distance is less computationally intensive. It emerges that the final sampling of the situations equivalent to the situation So observed has the advantage of being relatively fast by reducing the number of candidate equivalent situations. For each of the four observed flow rates, the observed occupancy rate, the observed average speed, and the expected demand, a measurement distance calculation is performed for each of the four variables to measure the difference between the situation. during the observed period, which is used before the forecast, with the comparable situations retained. For flow rates, the distance by partial measurements will be given by n summed differences for each of the n situations to be compared, these differences d (q;) being expressed by: d (qi) = n / Te * (Sum (I6gh, ohserved - The three other differences are expressed by relations of the same kind and are denoted d (to), d (v) and d (d) respectively for the rate. occupation, average speed and demand. The principle of distance by measurements consists, for each situation S '; retained in the previous step in the set Ep, to calculate for each magnitude of flow rate, occupancy rate, average speed and demand differences two by two lists of these quantities. Thus on the flow, one calculates a triangular matrix whose current term a;,; is: a ;; _ {q; - q; for h; <h ;, OR 0 for h; 5 h; ; These two-by-two differences are thus computed for n matrices associated with the rate q of n situations S '; retained during the step of calculating cumulative distances. The same calculations are made for the occupancy rate To, the average speed V and the request D. Thus, four measurement matrices are obtained for the compared situations belonging to the set Rp of the Np situations S '; retained during the cumulative distances step. The same calculation is then performed on the observed situation So and four measurement matrices are obtained from the four observed quantities of flow rate Q, occupancy rate To, average speed V and demand D, whose current term for the flow qo will be bi, j defined by: b;, i = {gio - q; o for hi <h ;, OR 0 for h; 5 hi; In another step, we calculate the profile difference between the last observed situation So and the compared situation S '; Among the n selected situations Ep. This step makes it possible to characterize the differences of shapes of the graphical representations of the histories of the four quantities of flow Q, of occupation rate To, of average speed V and of demand D. The difference of profile is characterized by a value obtained using an arithmetic function such that: - if the profile of the evaluated magnitude of the compared situation S '; and the profile of the magnitude evaluated for the observed situation So are identical, the value is 1; if the two compared profiles are different, the value is greater than 1 as the difference between the profiles is important. In one embodiment, this arithmetic function is obtained by the ratio of the medians of the set of observed values and the set of compared values, medianeo and medianeo, respectively. We will have the relation of definition of the difference of profile (p: 4) = medianeo / medianeo or the inverse relation so as to obtain a result superior or equal to 1 characterizing the difference. The profile deviations are thus calculated for each of the four flow variables Q, occupancy rate To, average speed V and demand D and are denoted 4) Q, 4) T0, 4) V and YI). In a next step, the arithmetic mean of the measurement differences is then calculated by summing the differences between the measurements on the four quantities considered for the compared situations S '; of Rp and for the observed situations So: mQ = ri / Te * (Sum (Ibhi, h; -ah ;,; ho - Te 5 h; <hi 5 ho) These averages are calculated for the four quantities of flow Q, occupancy rate To, average speed V and demand D. In a next step, the distance is calculated by measurements for each of the four flow variables Q, occupancy rate To, average speed V and demand. D. The relation of definition of the distance by measurement for the flow is: dMq; = ~ q * mq / Go; 1 5i5Np The distance by measurements between the situations compared If of the set Rp of the equivalent situations retained at previous step of selection of the Np equivalent situations of the observed situation So is computed in the form of the values: d mi = AQ * d MQ + ATo * d MTo + Av * d Mv + Ad * d Md In which relation the coefficients Ad, ATo, Av and Ad are selected by calibration of the very short-term forecasting method on test data, and in which dMX are the cadastral distances for the flow rate q, the occupancy rate T, the average speed V and the demand D. In one embodiment, the first three coefficients of weighting of the cumulative distance are equal to unity. , and the last is worth the sum of the first three, so as to obtain an equal weighting between the values corresponding to the explanatory base (Q, To and V) and the value corresponding to the explained basis (D). Finally, in a final step, the best equivalent situation is determined by looking for the lowest of the Np distances by measurement: io is the equivalence such that dM; o = min (dM;; 1 5 i 5 Np). In Figure 1, there is also shown an editing step (E5) of a forecast of a very short-term road traffic condition on the basis of the equivalent traffic situation in the form of a data set. planned traffic. In this step E5, the very short-term forecast is constructed from the historical I-day measurements that present the situation with the best equivalence with the situation observed on the last day of observation before the forecast period. Thus, the forecast includes measurements on the three quantities selected to characterize a traffic situation: the flow rate, the occupancy rate and the average speed of the vehicles in front of each measurement point 2; x curvilinear abscissa. By designating any of these three magnitudes by m, the prediction on the day jo from the horodate ho on the horizon Th is determined from the same time stamp at the beginning of the time slot for the same duration Th at the day I: mp, h = me, h; for ho <h <ho + Th and for all x; Finally, in a step E6, forecasting actions are performed that include: - the dissemination of road forecasting information; and control of traffic control actuators such as traffic lights; which will be described later using Figure 2.

The method of the invention furthermore comprises, in one embodiment, the E7 test of a condition for executing a loopback, that is to say for producing a new forecast in the very short term, for example by relaunching on the request of an operator the calculations E2 to E5 to a new horodate ho and if necessary for a predetermined sampling curvilinear abscissa x; on the road network 1, on which the forecast is made. In another embodiment, the loopback is executed periodically for example at a predetermined time of each day. The loopback is determined by a test in step E7. The test can be replaced by an infinite loop which allows the forecast to be renewed for each new measurement and for each measuring point of the road network. It should be noted that during a process start step E1, the initialization of the constant parameters used in the process described above, such as the prediction horizon Th or the number of situations to be sampled Np, are executed. These parameters can be initialized by a user interface, on a software run on a computer machine implementing the method of the invention, by a human operator, or by an adaptive process to achieve a constraint such as, for example, a time of calculation or prediction resolution, to adjust the accuracy and / or reliability of the forecast. In Figure 2, there is shown a block diagram of an embodiment of a traffic management system according to the invention. The road network 1 and the sensor group 2 have already been described with the traffic situation history database 10. A communication network 3 makes it possible to collect the measurement data from the sensors 2. It is connected to a module 4 of preparation of the data to be recorded in the database of traffic situation histories 10. The module 4 operates according to the teaching of step E2 of the method described above. Note that the data preparation module 4 can be partially distributed on the communications network 3 and / or on the various sensors 2 arranged at various curvilinear abscissae on the road network 1.

The system of the invention then comprises a module 5 for predicting a traffic request which implements the teaching of step E3 of the method of the invention. The module 5 is implemented on a computer arranged at a point of the communications network or on its own link with the data preparation module 4. The modules 4 and 5 as well as the database 10 can be implemented on a single computer or on a local area network connected to the communication network 3. The module 6 is an identification module of an equivalent traffic situation which implements step E4 of the process described above. Module 6 is implemented in the same way as the aforementioned modules. It takes its source data from the database 10, as well as the module 5 traffic demand forecast. A module 7 for editing a traffic forecast implements step E5 of the method described above. The module 7 is implemented in the same way as the aforementioned modules. It takes its source data from the database 10, as well as modules 5 for traffic demand forecasting and 6 for identifying an equivalent traffic situation. A traffic management module 8 then executes actions taken on the basis of the traffic forecast edited by the module 7. The module 8 comprises, in one embodiment, means for editing traffic forecast information bulletins. to terminals. These traffic information terminals are connected to it via a communication network which depends on the terminals used.

In one embodiment, the terminals are mobile phones or similar devices, and the communication network is a data network on the mobile network. In one embodiment, the terminals are onboard terminals on vehicles that plan to enter the road network 1. The communication network connecting them to the module 8 is then a mobile telephone network, the onboard terminal being adapted to such a mobile phone network. In another embodiment, the on-board terminal is a radio-frequency receiver for a vehicle and the communication network can then be a voice or digital type of traffic information network such as the RDS-TMC network. In another embodiment, the traffic management module 8 is connected to a network of traffic information displays, such as the display 9 shown in FIG. 2, and arranged at given abscissae on the road network 1. The traffic management module 8 then comprises a traffic information editor adapted to communicate with the displays 9. In another embodiment, the traffic management module 8 is connected by known means to a network of actuators traffic including, but not limited to, traffic lights, vehicle speed limitation signs, access barriers to one or more road lanes. In this embodiment, the traffic management module 8 includes means for generating control signals for controlling the traffic actuators. In another embodiment, the traffic management module 8 is connected by means of communication to toll barriers, so that the charging of the access or the exit of a vehicle depends in particular on the forecast of very short-term traffic generated by the traffic management system of the invention. For this purpose, the module 8 comprises means for determining a tariff control message intended for payment machines and / or for display on displays reserved for this purpose at each toll gate and intended for users and / or for staff responsible for collecting tolls. Other applications are possible within the scope of the invention. Likewise, the calculation methods indicated have been used in embodiments of the traffic management method according to the goals that have been defined. They are modifiable by those skilled in the art implementing the teaching of the invention without departing from the scope claimed. Other features of the invention have an importance that is emphasized here.

The step (E2) of preparing a set of data comprises at least one of the following steps: disqualification of the aberrant measurement data; - disqualification of the last measure preceding an outlier measurement data; - disqualification of the first measurement following an aberrant measurement data; smoothing the measurement data by calculating an average over a determined period of smoothing of the measurement data; - reconstitution of the missing measurement data by interpolation of the missing variables from the accompanying measurement variables. The forecasting step (E3) of the traffic request based on the measured traffic data comprises at least one of the following steps: determining a type of day according to a predetermined calendar typology; - determination of a nominal traffic demand profile for each type of day; - for an observed day, determination of the optimal nominal demand profile; - forecast nominal demand over a forecast period based on the optimal nominal demand profile; - adjustment of the nominal demand forecast. The editing step (E5) of a forecast of a road traffic state includes a step of calculating a forecast situation on the basis of the equivalent situation on the basis of at least one historical situation counted at from the timestamp of the equivalent situation taken within a given forecast period. The disqualification step of the aberrant measurement data involves testing at least one criterion, however, on: negative or zero flow rates; - strictly negative occupancy rates; and negative mean velocities, zero or greater than a threshold of disability of the measured velocities for disqualifying the measurement (A (t ;, x;)) at the time stamp (t;) and at the abscissa measuring point ( x;).

Especially in the latter case, the verification of only one of the criteria disqualifies the measurement (A (ti, xi). The step of reconstituting the missing data is preceded by a step of testing an interpolation condition of that the period between the last measurement preceding a period of missing data and the first measurement following said period of missing data is less than a predetermined period of interpolation according to a relation of the form: 0 <tb - ta <Ainterpol L The optimum nominal demand profile identification step involves comparing the set of nominal demand profiles established for each type of day noted: {{pc, h; h hd. <h <hF}; c in Ecalendar} flow profile of the observed day, calculated from the weighted flow rates {pio, h; hp, <h <ho} with the relation: pj0, h - gj0, h <qh> where <qh> is the mean of the flows observed on the period hp <h <ho of the last part to observe on the last day observed before the very short-term forecast period.

The step of identifying the optimal nominal demand profile comprises a step of calculating the current term dh of an expected demand d, given by the relation: dh-pi, h * <qh> where h is the current time stamp of the term calculated in the very short-term forecast period Ts; PHI is the h-th term of the optimal nominal demand profile selected in the previous step; and <qh> is the average of the flows on a predefined time slot, for the timestamps h between hp and hF of beginning and end of time slot. The optimal nominal demand profile identification step includes a processing step to fit the historical data.

In particular, the processing step for the identification of the demand profile is a resetting operation of the nominal demand forecast as a function of the flow rate observed over the explanatory period. In particular, the registration is performed by averages and / or medians.

The step of resetting the planned demand comprises the following successive operations - calculating the differences in flow rates observed two by two; - calculation of the deviations of expected requests two by two; - calculation of the ratio of the deviations of the requests to the deviations of the flows; Fmoyennecalculating the divergence of the ratio of deviations by mean and median Fmédiane, respectively by calculating the averages of the ratios of the deviations and the medians of the reports of the deviations; calculation allowing the scaling of the expected demand according to the divergences obtained previously, ie dh, 1 for the mean F average and dh, 2 for the median Fmedian, according to a relation of the form dh, 1 = dh / F average and dh, 2 = dh / Fmédiane, for all time stamps h, in the current time slot before the horodate h0 of the last traffic observation before the forecast period; calculating an offset p between the observed flow rates and the scaled demand in each of the p1 or median p2 cropping cases, according to the respective relations: p1 = sum (qh - dh, 1; hp < h <ho) and p2 = sum (qh - dh, 2; hp <h <ho); - calculation of adjustment of the volume of nominal demand forecast and recalculated according to the two mean and median methods according to the respective relations Dh, 1 = dh, 1 + p1 and Dh, 2 = dh, 2 + p2, for the time stamps h between hp <h <ho. The step of selecting the sample of the best equivalences consists in constituting an ordered set D of the distances dMi and dNi, sorted in ascending order; then, to select the Np first elements of D, set of pre-sampling observed situations equivalent to the situation So, with Np a predetermined number selected during the initialization step E1 of the method and thus to generate a set Rp of the references of Np situations equivalent to So.

Then a final sampling step of the situations equivalent to the last observed situation So is carried out using a distance function of better accuracy than the distance function used for the calculation of the set Rp of the Np references of the best situations equivalent to So. A distance function is used by measurement and the following steps are performed for this purpose: - calculation of distances by measurement; - selection of the best measurement.

The step of calculating distances by measurement comprises the following steps: - Calculation of the average of the distances by measurement; - Evaluation of the distance by measurement. The method of the invention also comprises a test step (E7) of a condition for performing a loopback, periodically executed for example at a predetermined time of each day, or by operator control. It can then comprise, in addition, a step (E1) of the beginning of the method, to execute the initialization of the constant parameters used by means of a user interface, on a software executed on a computer machine implementing the method of the invention, by a human operator, or an adaptive process to achieve a constraint such as a computation time or a prediction resolution, for adjusting the accuracy and / or reliability of the prediction. The system of the invention comprises a communication network (3) for collecting the measurement data of the traffic situation sensors (2) at specific time stamps connected to a module (4) for preparing the data to be recorded in the database. traffic history data (10). In one embodiment, it comprises a module (5) for predicting a traffic request. It comprises a module (6) for identifying an equivalent traffic situation, which takes its source data from the database (10), as well as the module (5) for predicting traffic demand.

It comprises a module (7) for editing a traffic forecast, which takes its source data from the database (10), as well as modules (5) for forecasting traffic demand and (6) for identification of an equivalent traffic situation.

It finally comprises a traffic management module (8), which then performs actions taken on the basis of the traffic forecast published by the module (7).

Claims (14)

REVENDICATIONS1. A method for managing road traffic, characterized in that it comprises the steps of: - Preparing (E2) a determined set of measured traffic data and storing the measured data in a history of traffic situations to be recorded in a base historical data (10); - Prediction (E3) of the traffic request on the history database (10); - Identification (E4) of an equivalent situation in the history of past traffic situations on the basis of the expected traffic demand; and - Editing (E5) a forecast of a very short-term road traffic condition based on the equivalent traffic situation in the form of a set of expected traffic data. 2. Method according to claim 1, characterized in that the step of identifying (E4) an equivalent traffic situation in a history of past traffic requests consists in: determining a traffic request identification period ; determining at least one equivalent situation in a situation history recorded in the situation history database, the traffic request identification period in the history of the observed requests. 3. Method according to claim 1, characterized in that the step (E3) for predicting the traffic demand includes a step of determining a nominal traffic demand profile for each type of day established on a predetermined calendar typology. the step of determining a demand profile comprising a step of determining: a set of averages of the flows over the entire history; - a set of weighted flows for the days of a given calendar typology, for each day of this typology on the basis of historical data according to a relation of the form: <qh> = average (q, h) ; j of calendar typology c and hp <h <hFor h is a time stamp in the time slot defined in each day type day c and mean () is a mean calculation function; a median of the weighted flows for each type of day of the typology determined according to a relation of the form: pc, h = median (p, h); for each day j belonging to the class c in the calendar typology Ecalendaire, and hp <h <hF, taken as nominal demand profile for the type c of day. 4. Method according to claim 3, characterized in that it comprises a step of selecting the expected nominal demand recaled by searching for the smallest distance between the observed flow rates Q and respectively the expected nominal demand D, the expected nominal demand recalibrated by a mean D, and the expected nominal demand recaled by a median D2, and that as a function of distance, a cumulative distance function is used, as sum of the partial distances each calculated as the distance between two elements of the same rank, according to a predetermined standard. 5. Method according to claim 4, characterized in that the step of identifying (E4) an equivalent situation in the history of past traffic situations on the basis of the expected nominal traffic demand comprises a step of: - pre-sampling of equivalent situations; then - sampling of equivalent situations based on pre-sampling. 6. Method according to claim 5, characterized in that the step of pre-sampling the equivalent situations consists in finding the smallest distance between candidate situations and the observed situation (So) during the last observation preceding the forecasting period. in the very short term, on the basis of the cumulative distance method, as the sum of the partial distances each calculated as the distance between two elements of the same rank, according to a predetermined norm, and applied to n situations Si having the same time slot in the as observed at the last observation preceding the very short-term forecast period, taken from at least one of three flow, occupancy and / or average speed and nominal traffic demand observed from situations close to the So situation, and previously subjected to a statistical treatment of a hundred rabies and reduction. 7. Method according to claim 6, characterized in that the step of sampling the equivalent situations consists of looking for the smallest distances between candidate situations taken from the situations obtained by pre-sampling and the observed situation (So) during the last observation preceding the very short-term forecast period, based on the measurement distance method, taken on at least one of three quantities of flow, occupancy and / or average speed, and of nominal traffic demand observed in situations close to the situation (So) and in that the editing step (E5) of a forecast of a very short-term road traffic situation on the basis of the situation of equivalent traffic in the form of a set of planned traffic data involves a very short-term forecasting construction step from the day I traffic measurements of the history that feel the situation with the smallest distance measurement from the situations identified in the sampling stage equivalent situations. 8. The method of claim 1, characterized in that it comprises a step (E6) for performing prediction actions which comprise: - the dissemination of road forecast information; and / or - the control of traffic control actuators as traffic lights; on the basis of the edition (E5) of a forecast of a very short-term road traffic situation. 9. Road traffic management system for implementing the method according to one of the preceding claims, characterized in that it comprises: - a database (10) of past traffic measurements; a set (2) of traffic data sensors to be measured; a means for processing (4) traffic measurement data; means for calculating a traffic request (5) at a given time stamp; - a means of identification (6) of an equivalent traffic situation; - a calculation means (7) of a forecast of a very short-term road traffic situation; and at least one operating device (9) for very short-term traffic condition forecast data. 10. System according to the preceding claim, characterized in that it comprises a module (8) for traffic management comprises means for editing traffic forecast information bulletins to terminals. 11. System according to the preceding claim, characterized in that the terminals are: - mobile phones or similar devices, and the communication network comprising a data network on a mobile network; and / or - onboard terminals on vehicles connected to the module (8) by a mobile telephone network, the onboard terminal being adapted to such a mobile telephone network as a radio-frequency receiver for a vehicle and the communications network comprising a network of voice or digital traffic information such as the RDS-TMC network. 12. System according to claim 10, characterized in that the module (8) for traffic management is connected to a network of traffic information displays, and then comprises a road information editor adapted to communicate with the displays. 13. System according to claim 10, characterized in that the module (8) for traffic management is connected to a network of traffic actuators comprising in particular traffic lights, vehicle speed limit panels, road barriers and traffic barriers. access to one or more roads of the road network, the traffic management module (8) comprising means for generating control signals for controlling the traffic actuators. 14. System according to claim 13, characterized in that the traffic management module (8) is connected by communication means to toll barriers, so that the charging of the access or the exit of a vehicle depends in particular on the forecast of very short-term traffic generated by the traffic management system of the invention, and in that the module (8) comprises means for determining a charging control message intended for the PLCs of the invention. payment and / or for display on displays reserved for this purpose at each tollgate and provided for the users and / or for the personnel charged with the collection of tolls.