Method for determining the flow of vehicles on a road network

Abstract

The present invention relates to a method for determining a vehicle flow rate on a road section of a road network. A model is constructed for determining a vehicle flow rate on a road section in the following manner: for each road section of a learning road network, an ordered sequence is determined according to a traffic direction comprising static descriptors of the learning road section and of its neighbors; a short-term and long-term memory neural network is trained on the ordered sequences and the result is concatenated with dynamic descriptors of the learning road sections; a transformer neural network is trained on these concatenations. Then the vehicle flow rate on a road section of a road network of interest is determined by applying the constructed model to static and dynamic descriptors of the road section and to static descriptors of the neighboring road sections. Figure 2A to be published

Description

Title of the invention: Method for determining a flow of vehicles on a road network Technical field

[0001] The present invention relates to the field of determining information relating to vehicle traffic for a road network. In particular, the invention relates to a method for determining a vehicle flow rate on at least one portion of road in a road network.

[0002] Today, metropolises and road managers have travel modeling tools to plan and simulate the impact of regulatory measures and future works, in order to reduce congestion or improve air quality. However, these models are very difficult to implement (need for very expensive population survey data, need for a significant calibration effort), difficult to maintain (maintenance is delegated to design offices, it is therefore difficult to develop the tool quickly), and difficult to update (population surveys are conducted every 5 to 10 years).

[0003] There is therefore a strong need for tools for determining vehicle flow rates within a transport network without having to use input data that is too costly and difficult to obtain, such as mobility surveys, which are also static and not suitable for predicting rapid changes in mobility, or real-life usage data, such as "floating car data" (FCD, which can be translated as data on journeys made and measured), which are dynamic but not necessarily available on all the road sections of an urban network. Furthermore, these alternative tools should be simple to use by non-experts, quick to execute to easily evaluate and compare several case studies, and finally reliable by reproducing recent counting data on the road network in question as best as possible. Prior art

[0004] The majority of alternative approaches in the literature require one or more measurements on a given road section to be able to estimate the corresponding traffic flow. Their accuracy depends on the penetration rate of the measurements and their availability. For these methods, it would therefore be impossible to estimate the flow on a road section that is not covered by these measurements. Some documents propose extrapolations and spatio-temporal correlations in order to estimate vehicle flows on a road network even without complete coverage of the measurements. They require continuous updating of the measurements and are often time-consuming. calculations and need for computer memory, with matrices that increase exponentially with the road network. Their use on a large road network can therefore be limited.

[0005] Vehicle flow prediction is a fundamental topic for the real-time operation and management of transportation systems. The application can range from road network monitoring, incident detection, or even emissions and air quality calculation. A large number of studies in the literature propose the estimation of traffic flow from counting loop measurements. They therefore exploit historical time series in order to predict traffic flow over a certain time horizon (temporal extrapolations). The models proposed to solve this problem are quite broad and vary from autoregressive statistical models to “machine learning” models such as neural networks, support vector machines (SVM), or random forests. “Deep learning” approaches have also been explored in the literature.These approaches often lead to better consideration of temporal correlations compared to machine learning approaches. However, these methods are limited to temporal predictions without necessarily studying spatial extrapolation. It is therefore impossible with these methods to predict flows on road sections without flow measurements.

[0006] Other methods have proposed estimating traffic flow from other sensors. A first method developed consists of estimating road traffic flow from data from surveillance videos. Other solutions involve statistical approaches and supervised learning approaches such as CNN (Convolutional Neural Network), GMM (Gaussian Mixture Model), or Kalman filters to estimate the number and direction of vehicles. Other methods are based on estimating road traffic flow from noise measurements from acoustic beacons. The approach consists of extracting characteristics from the acoustic signal and correlating them with road traffic measurements using supervised learning methods.However, the exploitation of the method requires the continuous presence of an acoustic signal with a sensor that is present next to a road segment to avoid background noise. Cellular data has also been used in the literature for the estimation of traffic flow. These types of approaches are however limited in accuracy, which depends on the penetration rate of the measurements and their availability. Also, it would be impossible to estimate the flow of vehicles on a road section that is not covered by these measurements.

[0007] Other authors have proposed methods for estimating road flow from various data and sensors present on a road network. Some proposed methods exploit FCD data and surveillance video data and / or measurement data from counting loops in order to estimate the flow on a complete road network. The approaches can range from semi-supervised graph-type learning to geometric matrix completions and spatial-temporal filters. This type of approach therefore makes it possible to estimate the flow on road strands even without measurement. This requires, however, a more or less significant penetration of measurements on the road strands with a continuous update of the measurements. These approaches are also often heavy in computation time with matrices that grow exponentially with the size of the road network. Their exploitation on a large road network can therefore be limited.

[0008] Data fusion techniques are also known, aimed at combining different sources of information, which may include sensor data, historical traffic data, weather data, etc. Finally, location data from smartphones (or smart phones to use the French term for them) and navigation applications are also used to estimate traffic density and predict traffic conditions.

[0009] Most of the aforementioned methods require at least one measurement on a specific road strand to estimate the associated traffic flow. The accuracy of these methods depends closely on the penetration rate of the available measurements and their availability. Therefore, these approaches prove incapable of estimating the traffic flow on strands not covered by these measurements (hereinafter referred to as spatial extrapolation).

[0010] However, document CN111292534A is known, which relates to a method for estimating traffic conditions for urban motorways, using k-means classification and deep sequence learning to estimate traffic conditions across the entire motorway network, even when real-time data is not available for certain road sections. The method involves dividing the motorway network, data modeling, data preprocessing, clustering analysis, and applying a deep sequence learning model to provide traffic information. However, this method has the disadvantage of being limited to motorway strands that form a network of long sequences that follow one another, unlike urban strands that have larger and more frequent intersections and bifurcations.

[0011] Also known is document CN 111222491A, which relates to a traffic evaluation system based on deep learning for estimating traffic flow using characteristic image sequences and convolutional neural networks. The paper aims to improve the detection of the number of vehicles based on the orthographic projection of their area in the images. This method concerns the processing of images obtained via cameras by deep learning. However, this method has the disadvantage of being dependent on the availability of images (via cameras for example) of the traffic to be able to deduce the flow. In addition, it is limited to the places where this information is available, unlike GIS (Geographic Information System) data which are available almost everywhere on the planet and easily accessible.

[0012] Document JP2023082282A2 is also known, which relates to a method for estimating travel time, capable of predicting the travel time required for a vehicle using a motorway to a predefined point using deep learning models. The proposed models first estimate the variations in traffic volume upstream in a continuous section of the motorway, and then calculate the travel time based on these predictions, also using output information to improve the accuracy of the predictions. This method, however, appears to be specific to motorways. Furthermore, the flow of vehicles is calculated only upstream of the section of the motorway used, in order to estimate a travel time.

[0013] Document IN202041039998A is also known, which relates to a method for predicting road flow, based on massive data to generate large-scale traffic data and which applies machine learning, image processing and deep learning algorithms to analyze this data, thus offering the possibility of predicting traffic flow and guiding autonomous vehicles for better traffic management in real time. The document then requires data coming in real time from mobile application users, which is then analyzed and combined with other information from autonomous vehicle sensors to better guide them. However, this method has the disadvantage of requiring real-time data (coming from applications such as Google Maps ™ for example), in addition to image / or video type data which are collected by autonomous vehicles.

[0014] Document CN101593424A is also known, which concerns a method for forecasting short-term traffic flow and which is based on three modules: a first historical average module, which divides the different dates of a year into three different types (working days, public holidays, and weekends) and which respectively calculates the different traffic flows for each type based on historical flow data; a second neural network module which estimates traffic flows using data modeling; finally, a third module which takes into account the different traffic conditions and combines the output of the historical average module and the neural network module to predict short-term traffic flow. Thus, this method allows the prediction of future flow rates on road strands, based on available flow measurements on the same strands. In other words, this method allows temporal extrapolation, but does not allow spatial extrapolation.

[0015] Document EP22204301A1 is also known, which relates to a method for determining a vehicle flow rate on a road section, including for a road section not equipped with a sensor. For this method, flow rate measurements are carried out, by at least one fixed sensor, at at least one measurement point of a learning road network. And a vehicle flow rate model is constructed by machine learning using macroscopic data from the learning road network and the measurements. This model is then applied to the road network section in question. This method makes it possible to quickly estimate (without traffic simulation tools) the traffic flow on a large road network in the absence of prior measurements. However, this method does not ensure the spatial continuity of the predicted flow rate between adjacent sections. Finally, this method is not sufficiently precise, as will be shown in the second application example below.

[0016] The present invention makes it possible to overcome these drawbacks. In particular, the present invention relates to a method for estimating a vehicle flow rate on any road strand of a road network and at any time. The method according to the invention is based on the combination of two deep neural networks, one taking into account spatial information from a learning road network, and the other taking into account temporal information on this same network. Furthermore, in the invention, the spatial information taken into account for deep learning relates to a road strand and the adjacent strands, which allows the neural network to learn spatial correlations between the characteristics of the strands and the associated road flow rate.Generally speaking, the method according to the invention makes it possible to increase the precision of vehicle flow estimates, to improve the spatial continuity of the predicted flow (in particular by taking into account adjacent road strands in “triplets” of strands), to speed up calculations and to reduce the memory occupation of the trained models. Summary of the invention

[0017] The present invention relates to a method for determining a flow rate of vehicles on at least one road section of a road network, by means of a learning base comprising, for each learning road section, a plurality of learning road sections of a learning road network, at least a vehicle flow measurement, at least one static descriptor, at least one dynamic descriptor, and at least one static descriptor for learning road strands of said learning road network having a node in common with said learning road strand.

[0018] The method according to the invention comprises at least the following steps:

[0019] A) We construct a model to determine a flow of vehicles on a road section of a road network from static and dynamic descriptors of said road strand of said road network by applying at least the following sub-steps:

[0020] I) for each learning road strand of said plurality of learning road strands, an ordered sequence of static descriptors relating to said learning road strand is determined as a function of said static descriptors of said learning road strand and of said static descriptors or of a statistic of said static descriptors of said learning road strands having a node in common with said learning road strand, said ordered sequence of static descriptors being ordered according to a predefined direction of circulation of said vehicles on said learning road strand;

[0021] II) a short and long term memory neural network is trained on said plurality of ordered sequences of static descriptors relating to said plurality of learning road strands;

[0022] III) for each learning road strand of said plurality of learning road strands, a result of said short and long term memory neural network for said learning road strand is concatenated with said dynamic descriptors of said learning road strand;

[0023] IV) a transformer neural network is trained on said plurality of concatenations of said result of said short-term and long-term memory neural network for said learning road strand with said dynamic descriptors of said learning road strand;

[0024] B) at least one static descriptor and at least one dynamic descriptor of said road strand of said road network are obtained, as well as at least one static descriptor for road strands having a node in common with said road strand;

[0025] C) by means of said at least one static descriptor and said at least one dynamic descriptor of said road strand of said road network, as well as said at least one static descriptor for said road strands having a node in common with said road strand, and by means of said model for determining a vehicle flow rate on a road strand of a road network as a function of static and dynamic descriptors of said strand of said road network, said vehicle flow rate is determined on said road strand of said road network.

[0026] According to an implementation of the invention, said learning base can be constructed by carrying out at least the following sub-steps for each learning road strand of said plurality of learning road strands of said learning road network:

[0027] i) At least one flow rate of vehicles on said learning road section is measured by means of at least one fixed traffic measurement sensor;

[0028] ii) At least one static descriptor and at least one dynamic descriptor are obtained for said learning road strand, as well as a static descriptor for each of said road strands having a common node with said learning road strand.

[0029] According to one implementation of the invention, said fixed traffic measurement sensor may be a counting loop.

[0030] According to an implementation of the invention, said static descriptors of said at least one road strand of said road network, respectively of said learning road strands of said learning road network, may comprise at least one type, one number of lanes, one length and one maximum speed of said road strand, respectively of said learning road strands.

[0031] According to an implementation of the invention, said dynamic descriptors of said at least one road strand of said road network, respectively of said learning road strands of said learning road network, may consist of a sequence of the average speeds of the vehicles for each hour of a day on said road strand, respectively on said learning road strands.

[0032] According to one implementation of the invention, said statistic may be an average.

[0033] According to one implementation of the invention, for said at least one road strand, respectively for each of said learning road strands of said learning road network, one can determine said ordered sequence of static descriptors relating to said road strand, respectively to said learning road strand, by forming a vector $ written: s=[ÿ, x, Z ]

[0034] where the vector x corresponds to said static descriptors of said road strand, respectively of said learning road strand, Y and Z are vectors of said statistics of said static descriptors of said road strands, respectively of said learning road strands, having a node in common with said road strand, respectively with said learning road strand, and located upstream and downstream of said road strand, respectively of said learning road strand, according to said predefined direction of circulation.

[0035] According to an implementation of the invention, for said at least one road strand, respectively for each of said learning road strands, it is possible to concatenate said result of said short and long term memory neural network for said at least one at least one road strand, respectively for said learning road strand, with said dynamic descriptors of said at least one road strand, respectively of said learning road strand, in the following manner:

[0036] ST = [h, T]

[0037] where is a vector corresponding to the output of said short-term and long-term memory neural network for said at least one road strand, respectively for said at least one learning road strand, T is a vector of said dynamic descriptors of said at least one road strand, respectively of said learning road strand, in the form T= (rt: t GT), where Tt is one of said dynamic descriptors of said at least one road strand, respectively of said learning road strand, measured at time at time step r of a measurement period F.

[0038] According to an implementation of the invention, it is possible to adapt infrastructures of said road network as a function of said vehicle flow rate determined for said at least one road section, and / or speed limits of said road network are adapted as a function of said vehicle flow rate determined for said at least one road section, and / or said vehicle flow rate determined for said at least one road section is displayed and / or an air quality estimated from said vehicle flow rate determined for said at least one road section is displayed.

[0039] Furthermore, the invention relates to a computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor, comprising program code instructions for implementing the method as described above, when said program is executed on a computer.

[0040] Other characteristics and advantages of the method according to the invention will appear on reading the following description of non-limiting examples of embodiments, with reference to the appended figures described below. List of figures [Fig IA] [Fig IB] [Fig IC]

[0041] [Fig.lA] (respectively [Fig.lB] and [Fig.lC]) presents a curve representative of the flow rate measured as a function of the time of day and a curve representative of the flow rate as a function of the time of day determined by the method according to the invention for a first (respectively a second and a third) road section of the validation database from the learning road network. [Fig 2A]

[0042] [Fig.2A] shows a map of the flow rate estimated by the method according to the invention on the main axes of the city of Lyon. [Fig 2B]

[0043] [Fig.2B] shows an enlargement of a portion of the urban area of ​​the map of [Fig.2A]. Description of the embodiments

[0044] The present invention relates to a method for determining a flow rate of vehicles on at least one section of a road network, preferably on the entire road network. In other words, the present invention aims to count a number of vehicles per unit of time which pass through a section of road.

[0045] According to one implementation of the invention, the vehicle flow rate on the road section considered may be, for example, an average flow rate or a maximum flow rate per unit of time, or more generally any statistic relating to a vehicle flow rate (minimum, standard deviation, median, etc.). The maximum vehicle flow rate reflects the capacity of a road section. This is a physical parameter which sizes the road section.

[0046] According to one implementation of the invention, the flow rate of vehicles on the road section considered may be a daily flow rate (for example average or maximum), i.e. a flow rate of vehicles on the road section considered for a duration of one day. This may make it possible to predict the variation in the flow rate from one day to another, in particular due to the day of the week, the month, public holidays, school holiday periods, etc. Alternatively, the flow rate of vehicles on the road section considered may be an hourly flow rate (for example average or maximum), i.e. a flow rate of vehicles on the road section considered for a duration of one hour. This may make it possible to predict the variation in the flow rate within a day, in particular due to work schedules, school schedules, leisure schedules, etc.

[0047] Generally speaking, a "road network" is a set of roads and paths in a predefined geographical area. This predefined geographical area can be a district of a city, a town, a community of communes, a department, etc. A strand of the road network is an elementary subdivision of the road network between two consecutive nodes of the road network. A road node is an end of a portion of road that can correspond to an intersection, an obstacle, a change in road signs, etc. For example, a strand of the road network can be a road between two consecutive intersections, between two consecutive signs, between an intersection and a sign, or a part of a motorway between two consecutive exits, etc. Thus, we have a fine division of the road network, and a model that is adapted to the road network without microscopic data.

[0048] In a conventional but non-limiting manner, the road network can be represented by a graph, called a road graph. The road graph is composed of a set of edges (also called arcs) and nodes, the nodes being able to represent the intersections, and the edges the portions of roads between the intersections. The graph can be obtained from an online mapping service ("webservice"), for example Here ™ (Here Apps LLC, Netherlands) which provides the edges of the graph as pure geometric objects. Preferably, the road graph is consistent with the road network (all physical connections between two roads, and only these, are represented by the nodes of the graph), as invariant as possible in time. In addition, the road graph can be simplified, by not taking into account portions of roads such as dead ends, paths in parks or cycle paths, depending on the type of the vehicle considered (for example for the embodiment of motor vehicles, cycle paths may not be taken into account).

[0049] The method according to the invention can be implemented for any type of vehicle: bicycle, motor vehicle, motorized two-wheeler, boat, hovercraft, scooter, etc. as well as pedestrian travel. However, the method according to the invention is particularly suitable for motorized vehicles, such as motor vehicles, heavy goods vehicles, buses, motorized two-wheelers, etc.

[0050] The method according to the invention requires measurements obtained by a fixed road traffic sensor. Each fixed road traffic sensor is arranged at a point (called a measurement point) of a road network (in this case a learning road network). Each fixed sensor allows punctual (spatially) and temporal measurements of the number of vehicle passages (vehicle flow) and possibly the speed of passage of the vehicles (traffic speed). The positions of the measurement points are identified within the road network. According to one embodiment of the invention, the fixed road traffic sensors can be cameras, radars, inductive loops, sensitive photoelectric cells detecting the break in a light beam, piezoelectric cables measuring the pressure exerted on the roadway or any similar sensors.

[0051] According to a preferred implementation of the invention, the fixed road traffic sensor corresponds to an inductive loop, also called a counting loop. There are in fact freely accessible databases of road flow measurements carried out by counting loops. The implementation of the method according to the invention is thus simplified.

[0052] The method for determining the flow rate of vehicles according to the invention can implement the following steps:

[0053] 1) Construction of the learning base

[0054] 1.1) Measurements of vehicle flow on a learning road network

[0055] 1.2) Obtaining static and dynamic descriptors of a plurality of strands road network learning drivers

[0056] 2) Construction of a model to determine a flow of vehicles on a strand of a road network using the learning base

[0057] 2.1) Determination of a plurality of ordered static descriptor sequences

[0058] 2.2) Training an LSTM neural network on the plurality of sequences static descriptor ordinates

[0059] 2.3) Concatenation of the result of the LSTM neural network with the descriptors dynamics

[0060] 2.4) Training a Transformer neural network on the concatenation of the result of the LSTM neural network with dynamic descriptors

[0061] 3) Determination of the flow of vehicles on a section of a road network using the model for determining vehicle flow on a section of a constructed road network

[0062] 3.1) Obtaining static and dynamic descriptors of at least one road strand of the road network

[0063] 3.2) Determination of ordered static descriptor sequences

[0064] 3.3) Application of the trained LSTM neural network

[0065] 3.4) Concatenation of the result of the LSTM neural network with the descriptors dynamics

[0066] 3.5) Application of the trained Transformer neural network on the concatenation of the result of the LSTM neural network with dynamic descriptors

[0067] Steps 1) and 2) can be carried out only once, and beforehand (“offline” mode). In other words, once the model for determining a vehicle flow rate on a section of a road network has been constructed using the learning base, this model can be applied one or more times. For example, if one wishes to determine the vehicle flow rate of several sections of a road network or the vehicle flow rate of the same road section but for different times, one can repeat only step 3 of the method according to the invention.

[0068] Advantageously, the learning base can be updated regularly (for example every month and at least every year), and generally as soon as new data is available, then the model can be updated to determine a flow rate of vehicles on a section of a road network, by training this model on the enriched learning base.

[0069] At least step 2) and / or step 3) can be implemented by computer means, in particular a computer, a processor or a calculator.

[0070] In the remainder of the application, the road network to which step 3) is applied is called the “road network considered” or “prediction road network”, or “inference road network”. It could also be called the “second road network”. This is the road network for which the vehicle flow rate is to be determined. These notations make it possible to distinguish it from the “learning road network”. ", or the "first road network", which is used to construct the learning base. The "road network considered" may correspond to or include the "learning road network", or be a road network different from the "learning road network".

[0071] The steps of the method according to the invention are detailed below. 1) Construction of the learning base

[0072] During this step, it is a question of constructing a learning base to train the model to determine a flow of vehicles on at least one strand of a road network.

[0073] This step comprises at least the two sub-steps described below.

[0074] 1.1) Measurements of vehicle flow on a learning road network

[0075] During this sub-step, it is a question of carrying out measurements of the flow of vehicles on a learning road network, using fixed road traffic sensors.

[0076] Very preferably, vehicle flow measurements are carried out at at least one point of a plurality of road strands of the learning road network. Conventionally, the plurality of input data of a neural network makes it possible to increase the capacity of the network to represent the reality to be modeled, and this is all the more true in deep learning. Preferably, the learning base comprises flow values ​​for at least 100 road strands of a learning road network, very preferably for at least 500 road strands of a learning road network.

[0077] Very preferably, the measurements of vehicular flow rate are repeated over time at a given point on each road section of a plurality of road sections of the learning road network, for example every hour, preferably every minute, of the same day. Such a periodicity of the measurement can make it possible to record the state of the traffic over time with precision, in order, for example, to determine peak hours and to observe the different fluctuations in flow rate during a day.

[0078] Advantageously, when the measurements are repeated over time for a period greater than one week, preferably one month, preferably one year, it is possible to associate with the measurement a categorical descriptor (for example a class representative of the day of the year) making it possible to categorize the measurements according to a given temporal segmentation, for example, working days, weekends, public holidays, etc. This can allow the neural network to learn to extrapolate temporally by integrating a categorical descriptor, specifying the chosen temporal segmentation. Thus, it will be possible to predict the flow rate for any day of the year.

[0079] 1.2) Obtaining static and dynamic descriptors of a plurality of strands road users of a learning road network

[0080] During this sub-step, it is a question of obtaining static and dynamic descriptors (or parameters, or attributes) relating to the plurality of road strands of the learning road network, as well as static descriptors for the road strands having a common node with the learning road strand. In other words, the learning base is constructed by taking into account static descriptors for the learning road strands, but also neighboring, or even adjacent, road strands. As will be described below, this contributes to ensuring spatial continuity of the predicted flow rate between adjacent strands.

[0081] By static descriptor relating to a strand of a road network (learning or not), we mean a parameter relating to the strand of the road network that is invariant over time. The static descriptors of the strands of a road network according to the invention can also be called “macroscopic data of the road network” or “structural data” in the literature. The macroscopic data correspond to the information related to the road network, such as the infrastructure, the slope, the signage, etc. As will be described below, in the invention, the static descriptors relating to a road strand (learning or not) are taken into account, as well as the static descriptors relating to its directly neighboring strands (having at least one node in common).

[0082] According to one aspect of the invention, the static descriptors of a strand of a road network (learning or not) may comprise the topography (i.e., slope, length, turns, intersections, etc.) of the road strand, the type of road strand (e.g., urban road, highway, etc.), the number of lanes of the road strand, the maximum speed of the road strand, the signage of the road strand, etc.

[0083] Preferably, the static descriptors of a strand of a road network (learning or not) comprise at least the type of road strand, the number of lanes of the road strand, the maximum speed of the road strand and the length of the road strand.

[0084] Preferably, the static descriptors of strands of a road network (learning or not) can be obtained by a geographic information system (GIS). Here ™ , Google Maps ™, OpenStreetMap ™ are examples of geographic information systems. In general, GIS propose a modeling of a road network in the form of a graph, where each strand is considered as a link between two nodes whose static descriptors are invariable. This representation is particularly suitable for the implementation of the invention, since one considers, for each strand of a road network, its neighboring (or even adjacent) strands, upstream and downstream, as will be described for the purposes of the invention below.

[0085] By dynamic descriptor (or temporal descriptor or time-varying descriptor) relating to a strand of a road network (learning or not), we mean a parameter relating to the road network strand varying over time. This can be a time series giving the average speed of vehicles on the road strand considered for each hour of a day. Such information can also be obtained via GIS.

[0086] As will be described below, the method according to the invention separately exploits the static and dynamic descriptors according to the invention, or in other words, the method according to the invention implements a learning architecture separating the static descriptors from the dynamic descriptors. The separation of the static and dynamic descriptors makes it possible to use models (in this case recurrent neural networks) that are different and adapted to their specific characteristics. For example, as will be described below, so-called “attention” models are more suited to keeping a memory of long temporal sequences, and therefore will be more suited to dynamic descriptors. Similarly, as will be described below, LSTM-type neural networks make it possible to learn the spatial dependencies of relatively short sequences of static descriptors.

[0087] Thus, at the end of this step, the learning base according to the invention comprises at least static and dynamic descriptors for a plurality of road strands of a learning road network, as well as the static descriptors of the strands having a node in common with the road strands of the learning road network.

[0088] Advantageously, and for the purposes of step 2) of building the model, the learning base can be divided into 2, preferably 3, subsets:

[0089] - a first subset, comprising for example between 70% and 85%, of preferably 80% of the entire training database data set, intended for training neural networks: this data will directly determine the estimation error measurement of the networks and how to adjust the values ​​of the neural network parameters; this is the training database ("training set" in English).

[0090] - a second subset, comprising for example between 15% and 30%, of preferably 20% of the entire training database data set, intended to validate the training process during its execution: this data is used to monitor the success of the training operation during its execution but does not directly influence the adjustment of the network parameters; this is the validation database. If the results obtained are not satisfactory, this may mean that the training database is not sufficiently representative and that it may be necessary to increase the training database with training data covering a wider range of road flows, by adding road strands.

[0091] - optionally but advantageously, a third subset, comprising a number of data from the learning base in a proportion equivalent to the validation database, intended to test the learning performance, at the end of the learning; it is the test database.

[0092] 2) Construction of a model to determine a flow of vehicles on a strand of a road network using the learning base

[0093] During this step, it is a question of constructing a model to determine a flow rate of vehicles on a strand of a road network from static and dynamic descriptors of the strand of the road network, by training, on the learning base constructed in step 1), a combination of two recurrent neural networks, composed of a short and long term memory neural network (known by the acronym LSTM, for “Long and Short Term Memory” in English) followed by a transformer neural network (“Transformer” in English).

[0094] The LSTM type network is part of the family of recurrent neural networks (RNN for "Recursive Neural Network" in English). A recurrent neural network is a network of artificial neurons presenting recurrent connections. A recurrent neural network is made up of interconnected units (neurons) interacting non-linearly and for which there is at least one cycle in the structure. The units are connected by arcs (synapses) which have a weight. The output of a neuron is a non-linear combination of its inputs. Thus, RNNs make it possible to follow a sequence of information. In particular, these networks are particularly suited to sequential data, because they use an internal memory to store information on previous time steps. However, recurrent neural networks face the problem of gradient disappearance when learning to memorize past events.

[0095] LSTM neural networks are improved variants of RNNs that address the gradient disappearance problem. Indeed, each computational unit is linked not only to a hidden state but also to a state of the cell that acts as a memory. The transition is done by transfer with constant gain equal to 1. In this way, errors propagate to previous steps without gradient disappearance phenomenon. The state of the cell can be modified through an input gate, which allows or blocks the update ("input gate"). Similarly, an output gate controls whether the cell state is communicated at the output of the LSTM unit ("output gate"). The most widespread version of LSTM also uses a gate allowing the cell state to be reset to zero ("forget gate").

[0096] Transformer network is also a variant of recurrent neural networks. Transformer networks are suitable for handling sequential data. However, unlike other RNNs, transformers do not require that sequential data be processed in order. Thanks to this feature, the transformer allows for much parallelization. greater than RNNs and therefore reduced training times. Chained RNNs process tokens sequentially, maintaining a state vector that contains a representation of the data seen after each token. To process the nth token, the model combines the state representing the input sequence up to the (nl)th token with the information from the new token to create a new state representing the input sequence up to the nth token. Theoretically, the information from a token can propagate far back to the beginning of the input sequence, if at each point the state continues to encode information about the token. But in practice, this mechanism is imperfect: because of the gradient vanishing problem, the state of the model at the end of processing a long input sequence cannot remember precise information about the first few tokens.To address this problem, transformer networks introduce attention mechanisms. These mechanisms allow a model to directly look at, and extract, the state of any previous token in the sequence. The attention layer can access all previous states and give them weight according to their relevance to the current token, thus providing salient information about distant tokens.

[0097] The nesting according to the invention of the two aforementioned recurrent neural networks is carried out according to steps 2.1 to 2.4 described below.

[0098] 2.1) Determination of a plurality of sequences of static descriptors ordered

[0099] During this sub-step, for each road strand of the plurality of road strands of the learning road network, an ordered sequence of static descriptors relating to the learning road strand is determined as a function of the static descriptors of the learning road strand and the static descriptors or a statistic of the static descriptors of the learning road strands having a node in common with the learning road strand, the ordered sequence of static descriptors being ordered according to a predefined direction of circulation of the vehicles on the learning road strand.

[0100] In other words, according to the invention, the ordered sequence of static descriptors comprises the static descriptors or a statistic of the static descriptors of the learning road strands having a node in common with the road strand considered and located upstream of the strand considered according to the predefined direction of circulation, followed by the static descriptors of the road strand considered, themselves followed by the static descriptors or the statistic of the static descriptors of the road strands having a node in common with the road strand considered and located downstream of the strand considered according to the predefined direction of circulation.

[0101] Such an ordered sequence makes it possible to take into account the static descriptors of the road strands adjacent or even neighboring the road strand considered, which is more in a direction of circulation.

[0102] According to a preferred implementation of the invention, the statistics of the static descriptors of the road strands having a node in common with the road strand considered can be an average. This makes it possible to take into account average trends of the static descriptors on the adjacent road strands.

[0103] The sequence of static descriptors ordered according to the invention associated with a road strand can be represented in the form of a vector. More precisely, consider a road strand having static descriptors represented by a vector xg P^- Let Ye pÆ xd be matrices relating to the k and k strands (k and k' can be greater than or equal to 1) located upstream and downstream (relative to the direction of circulation) of strand x, each upstream and downstream strand having d descriptors respectively.

[0104] According to an embodiment according to which the static descriptors of the strands adjacent to the strand considered are taken into account in the ordered sequence of static descriptors according to the invention in the form of a statistic (for example an average), the vector x can be concatenated with a vector Y e P^ and a vector ZG respectively comprising the statistic (by type of descriptor) of the static descriptors of the matrices Y and Z. We obtain a sequence (represented in the form of a vector) ordered ,$ e pw, called a triplet hereinafter, which can be written:

[0105] s = z]-

[0106] Thus, in this design, the vector $ represents an ordered sequence of static descriptors, noted subsequently, relating to a road strand, such that S = (sp: pe 3 xd )•

[0107] According to an implementation mode where each of the static descriptors of the strands adjacent to the strand considered are taken into account in the ordered sequence of static descriptors according to the invention, the ordered sequence according to the invention can be written:

[0108] S = [nf ...,Y1K, .... YDL .... YDK, x,Zlf ....ZUT, ZDl ... ; , ZDK'] with D varying from 1 to d, K varying from 1 to k, and K' varying from 1 to k', and $ ep(*+*'+i)xd. Thus, in this design, the vector 5' represents an ordered sequence of static descriptors, noted sp hereinafter, relating to a road strand, such that S = Gp ; p & (k + k + 1) xd )•

[0109] At the end of this sub-step, an ordered sequence of static descriptors is obtained for each road strand of the plurality of road strands of the learning road network according to the invention.

[0110] 2.2) Training an LSTM neural network on the plurality of sequences static descriptor ordinates

[0111] During this sub-step, an LSTM neural network is trained on the plurality of ordered sequences of static descriptors determined in the previous sub-step.

[0112] Subsequently, the following notation is used for the activation vectors of the linked LSTM neural network: - at the entrance door: ip; - at the exit door: °p; - at the gate of oblivion: fp; - to the state of the cell: cp; - at the exit door: hp.

[0113] According to an implementation of the invention, the activation vectors of the LSTM neural network, for each static descriptor sp, can be written as follows:

[0114] ip= <7( Wi - [hp.^ fy)

[0115] Op= &(WO - [hp.1, + bo)

[0116] fp = a(Wr [hp.b + bf)

[0117] cp= fp-cp^ + ip• tanh(wc• [h^, + b^

[0118] _ 0p.tanh(cp)

[0119] such that p is the spatial step (or the descriptor index) in the triplet -S, y, and VE c are weight vectors relative respectively to the input gate ip, to the output gate °p, to the forget gate fp , and to the state of the cell cp, and b^ bo, bf, and bc are bias vectors relative respectively to the input gate ip, to the output gate °p, to the forget gate fp , and to the state of the cell cp, and where 17 is a sigmoid function defined by . The weight vectors and the bias vectors are learned during the training process.

[0120] The above equations are valid only for a single spatial step. Thus, if the triplet is a sequence of N elements, these equations are calculated N times.

[0121] 2.3) Concatenation of the result of the LSTM neural network with the dynamic descriptors

[0122] During this sub-step, for each road strand of the plurality of road networks of the learning base, the result (or the output) of the LSTM determined in step 2.2) for the road strand considered is concatenated with the dynamic descriptors of this road strand.

[0123] In other words, during this step, the output h of the LSTM determined for the ordered sequence of static descriptors S of a strand x is concatenated with the dynamic descriptors of the strand x , denoted T - (rt: te P), according to the following formula:

[0124] ST = [h, T]

[0125] such that T is the measurement period of the dynamic descriptors considered (for example a duration of 24 hours).

[0126] 2.4) Training a Transformer neural network on concatenation of the result of the LSTM neural network with dynamic descriptors

[0127] During this sub-step, the transformer neural network is trained on the concatenation of the result of the LSTM neural network with the dynamic descriptors of each road strand, hereinafter called the "input of the Transformer". Conventionally, this means that the input, which we denote by ST, of the Transformer is introduced into the encoder part of the transformer.

[0128] According to an implementation of the invention, the encoder may be composed of a stack of identical encoding layers with the same series of operations. Each encoder may mainly contain a multi-head attention layer, a positional encoding layer and a linear layer connected with an activation function of the ReLU type (from "REctified Linear Unit" in English or rectified linear unit).

[0129] According to one implementation, in the attention layer, the transformer can learn three weight matrices: the query weight matrix Wq, the key matrix called attention head. These weight matrices contain parameters that are optimized during training. These parameters are used to weight the input values. The attention, denoted A, for the entire sequence can be expressed in matrix representation for the entire ST sequence as follows:

[0130] n (QKT\ K, V) = softmax -7^ . V \ /

[0131] where qe pwx^ and yep* x each line of the keys of the Q queries and U values ​​is a vector representation of sequence elements, and softmax is a function to convert a vector of real numbers into a probability distribution. The softmax function is used to normalize attention scores, ensuring that the attention given to each component of the sequence input to the transformer is proportional to its relative importance compared to other components in the sequence.

[0132] According to an implementation of the invention, during the training of the transformer, the ST sequence composed of static descriptors and dynamic descriptors k and the weight matrix of the values ​​Wy, 1 set (yy jy ST = / \ is transmitted all at once in the first Vf ^2' ......^3xd+r / encoder block, and then passed on to its successor, that is, to the next encoder. The process is repeated until all N encoding blocks have processed the transformer input. An encoded representation of ST is then obtained. In this context, models based on recurrent neural networks (RNNs) are limited in their ability to use information from elements observed far in the past in the sequence. More generally, they have more difficulty relating sequential information that is far apart. Attention, on the other hand, can efficiently and quickly relate each element in one sequence to all the elements in another sequence, as well as to all the other elements in the same sequence.

[0133] Thus, at the end of this step, a model is obtained for determining a flow rate of vehicles on a strand of a road network, from static and dynamic descriptors of the strand of the road network, the model comprising a combination of two recurrent neural networks, precisely an LSTM and a Transformer, trained on a learning basis.

[0134] 3) Determination of the flow of vehicles on at least one section of a road network using the model to determine a vehicle flow rate on a section of a constructed road network

[0135] During this step, the model constructed on a learning road network as described in step 2) is applied to at least one road section of a second road network, so as to determine the flow of vehicles on this road section.

[0136] More precisely, during this step, the following sub-steps are carried out:

[0137] 3.1) Obtaining static and dynamic descriptors of at least one strand road network road

[0138] During this sub-step, it is a question of obtaining static and dynamic descriptors relating to said at least one road strand of the road network considered, as well as static descriptors relating to the road strands of the second road network having a common node with said at least one road strand considered and being located upstream and / or downstream (with respect to a predefined direction of circulation) of said at least one road strand considered.

[0139] It is quite clear that this sub-step corresponds in every respect to sub-step 1.2) described above, except that it is applied here to said at least one road strand of the road network considered and to its adjacent strands, and no longer to a plurality of road strands of the learning road network and to their adjacent strands.

[0140] 3.2) Determination of ordered static descriptor sequences

[0141] During this sub-step, for the at least one road strand of the road network considered, an ordered sequence of static descriptors is determined, said ordered sequence of static descriptors being ordered according to a direction of circulation of said vehicles on said road strand, said ordered sequence of static descriptors comprising an average of the static descriptors of the road strands having a node in common with the road strand considered and located upstream of the strand considered according to said predefined direction of circulation, followed by the static descriptors of the road strand considered, followed by a statistic (for example an average) of the static descriptors of the road strands having a node in common with the road strand considered and located downstream of the strand considered according to said predefined direction of circulation.

[0142] It is quite clear that this sub-step corresponds in every respect to sub-step 2.1) described above, except that it is applied to said at least one road strand of the road network considered, and no longer to a plurality of road strands of the learning road network.

[0143] 3.3) Application of the trained LSTM neural network

[0144] During this sub-step, for at least one road strand of the road network considered, the LSTM neural network trained in step 2.2) is applied to at least the ordered sequence of static descriptors of the road strand determined in step 3.2).

[0145] 3.4) Concatenation of the result of the LSTM neural network with the dynamic descriptors

[0146] During this sub-step, for said at least one road strand of the road network considered, the result (or the output) of the LSTM determined in step 3.3) for the road strand considered is concatenated with the dynamic descriptors of this road strand.

[0147] It is quite clear that this sub-step corresponds in every respect to sub-step 2.3) described above, except that it is applied to said at least one road strand of the road network considered, and no longer to the plurality of road strands of the learning road network.

[0148] 3.5) Application of the trained Transformer neural network on the concatenation of the result of the LSTM neural network with the dynamic descriptors

[0149] During this sub-step, the transformer neural network is applied to the concatenation of the result of the LSTM neural network with the dynamic descriptors of said at least one road strand.

[0150] Thus, the method according to the invention is based on taking into account the road strands and their direct vicinity, which is moreover according to a direction of circulation (upstream and downstream strands). This description allows an increase in input data by a factor of 3, and an incorporation of the direction of travel in the structure of the descriptors. In addition, This allows us to take into account the fact that the road strands are intrinsically linked to each other. Indeed, the traffic state of a road strand is dependent on the traffic state of the neighboring strands located upstream and downstream. In addition, this allows us to take into account the fact that two strands with the same static descriptors will not necessarily present the same variations in flow.

[0151] Furthermore, the use of the LSTM as a model for extracting the spatial characteristics of a road strand obtained via the static descriptors makes it possible to consider the sequence of the triplet as a static series and therefore to preserve the continuity of the flow between the adjacent strands. To predict the time series representing the flow, the output of the LSTM (containing a dense representation of the spatial characteristics respecting the order of the strands) is concatenated with the dynamic descriptors. This creates a sequence twice as large as the output of the LSTM. This sequence then constitutes the input of the transformer, which makes it possible to capture long-term patterns. Its output can be followed by linear layers and ReLU activation functions allowing the flow to be predicted.Thus, the deep learning method implemented in the method according to the invention allows, via an LSTM, to capture local characteristics, and via a Transformer to capture long-term patterns. Furthermore, the present invention allows to have flow estimates for any territory without prior measurements made on this territory.

[0152] According to an implementation of the invention, from the flow of vehicles on at least one section of a determined road network as described above, it is possible: - adapt the road network infrastructure according to the said vehicle flow rate determined for the said at least one road section, for example by creating new roads, modifying the signage, etc. Based on the flow rate, public authorities and public works companies can in fact determine the roads with a high vehicle flow rate and adapt the roads to the users; - and / or adapt the speed limits of the road network according to said vehicle flow determined for said at least one road section, for example via illuminated signs according to the estimated flow downstream of said signs, in order to improve traffic flow; - and / or display said vehicle flow rate determined for said at least one road section or a parameter determined from said vehicle flow rate such as air quality, for example via a display on the dashboard of vehicles using the road network, on a stand-alone portable device, such as a geolocation device (GPS type), a mobile phone (smartphone type), on a website, etc. The display may take the form of a note or a color code or a representation thickness. of the road. A user of the road network may, depending on the predicted flow rate for the road section considered, choose to take a route not including this road section; - and / or display an estimated air quality based on the flow rate determined as described above, for example via a connected object (a panel) in a street or in a group of streets (neighborhood). Information on the vehicular flow rate of a road section can in fact allow an estimation of the air quality near this road section.

[0153] Thus, the present invention can be seen as a virtual vehicular flow sensor, in the sense that it makes it possible to simulate a vehicular flow measurement. It is thus possible to deploy a practically unlimited number of these virtual sensors at low cost to have an overall view of the flow over a territory.

[0154] Furthermore, the invention relates to a computer program product downloadable from a communication network and / or recorded on a medium readable by a computer (for example an on-board computer) and / or executable by a processor. This program comprises program code instructions for implementing the method as described above, in particular steps 2) and / or 3) described above, when the program is executed on a computer. Examples

[0155] The characteristics and advantages of the method according to the invention will appear more clearly on reading the application example below.

[0156] A learning base of road strands for which the road flows are known was used. For the construction of the model according to the invention, 70% of the strands are used to train the model according to the invention and the remaining 30% are used for validation. First example#

[0157] The first example consists of a comparison between the actual hourly flow rate (measured) and the flow rate estimated by the method according to the invention for three road sections of the validation database. Figure 1A (respectively Figure 1B and Figure IC) shows a MES curve representative of the flow rate D (in number of vehicles per hour, noted veh!h ) measured as a function of the hour H of the day (in hours) and an INV curve representative of the flow rate D (in number of vehicles per hour, noted veh / h ) as a function of the hour H of the day (in hours) determined by the method according to the invention for a first (respectively a second and a third) road section of the validation database from the learning road network. It can be observed in these figures that the model according to the invention is capable of predicting a large range of flow rate values ​​(values ​​between 300 - 900 veh!h).

[0158] In addition, the INV curves determined by the method according to the invention follow the trends of the measured MES curves, thus making it possible to predict the moments of congestion corresponding to the morning and evening peak hours (around 8 a.m. and 6 p.m.). Second example# The second example consists of a comparison of the flow rate determined by the method according to the prior art described in the aforementioned document EP22204301A1 and by the method according to the invention. Table 1 presents a comparison of the absolute error, the mean of the GEH statistic, the median value of the GEH statistic, as well as the percentage of GEH greater than 5, on the one hand determined by the method according to the invention (column INV) and on the other hand by the method according to the prior art (AA). The GEH statistic (from the initials of its creator Geoffrey E. Havers) is a formula used in traffic engineering (traffic forecasting and modeling in particular) to compare two volumes of traffic (typically, a series of actual measured traffic and a series of traffic resulting from modeling). It can be written:

[0159]

[0160] ^“GEH y j+y Where are the actual measurements and y are the values ​​estimated by a process. It is classically considered that if _ ^veh / hj and mo'ns ^es flow rates predicted have a GEH greater than $ then the modeled flow values are considered accurate. It can be observed that the method according to the invention has made it possible to considerably reduce the values ​​of these quantities, thus demonstrating the superior extrapolation capabilities of deep learning with regard to flow prediction. In particular, the value of the absolute error has been reduced by a factor greater than 3, the average GEH is less than 5 and therefore meets the conditions, and finally the % of GEH greater than 5 has been halved. [Tables 1] INV AA Absolute error (yeh / h) 59.67 203.4 Mean GEH 4.48 8.8 Median GEH 3.06 7.1 % of GEH > 5 30.52 63.1 Third example In the third example, the method according to the invention is applied to the city of Lyon (France) for a working day, at a peak hour (6 p.m.). The input data

[0161] of the model (static and dynamic descriptors) come from GIS services such as Here Maps. [Fig.2A] shows a map of the flow rate D (in number of vehicles per hour, represented in grayscale, the unit of which is noted veh / h) estimated by the method according to the invention on the main axes of the city of Lyon, and [Fig.2B] shows an enlargement of a part located in the urban area of ​​the map of [Fig.2A]. It can be observed in these figures that the peripheral strands (i.e. the strands located on the outskirts of the city) have a significant flow rate compared to the urban strands (i.e. the strands located in the center of the city). This corresponds well to the observed reality of road traffic. In addition, it can be observed that the spatial continuity of the flow rate is preserved. Indeed, it can be observed that the order of magnitude of the flow rate remains the same for consecutive strands where there is no crossroads or intersection.In addition, the flow rate prediction by the method according to the invention is obtained in approximately 1 / 4h on an Intel Core i7 @2.60 GHz processor with 16 GB of memory, i.e. a duration divided by two compared to the method according to the aforementioned prior art.

Claims

1. Claims Method for determining a vehicle flow rate on at least one road section of a road network, characterized in that at least the following steps are carried out by means of a learning base comprising, for each learning road section, a plurality of learning road sections of a learning road network, at least one measurement of the vehicle flow rate, at least one static descriptor, at least one dynamic descriptor, as well as at least one static descriptor for learning road sections of said learning road network having a node in common with said learning road section: A) A model is constructed to determine a vehicle flow rate on a road section of a road network from static and dynamic descriptors of said road section of said road network by applying at least the following sub-steps: I) for each learning road strand of said plurality of learning road strands, an ordered sequence of static descriptors relating to said learning road strand is determined as a function of said static descriptors of said learning road strand and of said static descriptors or of a statistic of said static descriptors of said learning road strands having a node in common with said learning road strand, said ordered sequence of static descriptors being ordered according to a predefined direction of circulation of said vehicles on said learning road strand; II) a short and long term memory neural network is trained on said plurality of ordered sequences of static descriptors relating to said plurality of learning road strands; III) for each learning road strand of said plurality of learning road strands, a result of said short-term and long-term memory neural network for said learning road strand is concatenated with said dynamic descriptors of said learning road strand; IV) a transformer neural network is trained on said plurality of concatenations of said result of said short-term and long-term memory neural network for said road strand learning with said dynamic descriptors of said learning road strand; B) at least one static descriptor and at least one dynamic descriptor of said road strand of said road network are obtained, as well as at least one static descriptor for road strands having a node in common with said road strand; C) by means of said at least one static descriptor and said at least one dynamic descriptor of said road strand of said road network, as well as said at least one static descriptor for said road strands having a node in common with said road strand, and by means of said model for determining a vehicle flow rate on a road strand of a road network as a function of static and dynamic descriptors of said strand of said road network, said vehicle flow rate is determined on said road strand of said road network.

2. Method according to claim 1, wherein said learning base is constructed by carrying out at least the following sub-steps for each learning road strand of said plurality of learning road strands of said learning road network: i. At least one vehicle flow rate is measured on said learning road strand by means of at least one fixed traffic measurement sensor; ii. At least one static descriptor and at least one dynamic descriptor of said learning road strand are obtained, as well as a static descriptor for each of said road strands having a common node with said learning road strand.

3. The method of claim 2, wherein said fixed traffic measurement sensor is a counting loop.

4. Method according to one of the preceding claims, wherein said static descriptors of said at least one road strand of said road network, respectively of said learning road strands of said learning road network, comprise at least one type, one number of lanes, one length and one maximum speed of said road strand, respectively of said learning road strands.

5. Method according to one of the preceding claims, in which said dynamic descriptors of said at least one road strand of said road network, respectively of said learning road strands of said learning road network, consist of a sequence of average vehicle speeds for each hour of a day on said road section, respectively on said learning road sections.

6. A method according to any preceding claim, wherein said statistic is an average.

7. Method according to one of the preceding claims, in which, for said at least one road strand, respectively for each of said learning road strands of said learning road network, said ordered sequence of static descriptors relating to said road strand, respectively to said learning road strand, is determined by forming a vector $ written: S = XZ] Where the vector x corresponds to said static descriptors of said road strand, respectively to said learning road strand, Y and Z are vectors of said statistics of said static descriptors of said road strands, respectively to said learning road strands, having a node in common with said road strand, respectively with said learning road strand, and located upstream and downstream of said road strand, respectively to said learning road strand, according to said predefined direction of circulation.

8. Method according to one of the preceding claims, in which, for said at least one road strand, respectively for each of said learning road strands, said result of said short-term and long-term memory neural network is concatenated for said at least one road strand, respectively for said learning road strand, with said dynamic descriptors of said at least one road strand, respectively of said learning road strand, in the following manner: ST = [A, T] where h is a vector corresponding to the output of said short-term and long-term memory neural network for said at least one road strand, respectively for said at least one learning road strand, T is a vector of said dynamic descriptors of said at least one road strand, respectively of said learning road strand, in the form T = ( vt : te F ) where Tt is one of said dynamic descriptors of said at least one road strand, respectively of said road strand learning road, measured at time step * of a measurement period T.

9. Computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor, comprising program code instructions for implementing the method according to one of the preceding claims, when said program is executed on a computer.

10. Method according to one of claims 1 to 8, in which road network infrastructure is adapted as a function of said vehicle flow rate determined for said at least one road section, and / or speed limits of said road network are adapted as a function of said vehicle flow rate determined for said at least one road section, and / or said vehicle flow rate determined for said at least one road section is displayed and / or an air quality estimated from said vehicle flow rate determined for said at least one road section is displayed.

Images