Method for determining at least one attribute of road traffic for a transport network

Abstract

The present invention relates to a method for determining at least one road traffic attribute (ATT) from at least two representations (RES A, RES B) of the transport network, at least one road traffic parameter being associated with one of the two representations (RES A, RES B) of the transport network. To determine the road traffic attribute, a third representation (RES C) of the transport network is constructed by means of a correspondence (COR) between the two representations of the transport network, the correspondence being implemented after a strand elimination step (ELI). Figure 1 to be published

Description

Title of the invention: Method for determining at least one attribute of road traffic for a transport network. Technical field

[0001] The present invention relates to the field of determining technical information concerning road vehicle traffic for a transport network, in particular a road network. Specifically, the invention relates to a method for determining vehicle flow and / or vehicle speed and / or the quantity of pollutants emitted by vehicles on a transport network, in particular a road network.

[0002] Having effective means of exchanging spatial information between different individuals has been a subject of paramount importance to humanity for many centuries. The science that deals with the modeling of spatial information is known as cartography. More precisely, cartography studies the graphic representation of a geographic area.

[0003] Over the past few decades, the explosion of technological advances in computer systems has revolutionized the field of cartography. This revolution has been driven primarily by the development of geographic information science, a term coined by Roger Tomlinson in the 1960s. A geographic information system (GIS) is a computer system that captures, stores, verifies, and displays data relating to locations on the Earth's surface. This technology is layer-based, meaning that any type of data containing location information can be added as a map layer to complete it. For example, layers can include socio-economic data about a population, landscape information, the locations of all types of sites (farms, factories, schools, shops, etc.), and, notably, road networks.

[0004] The main objective of a geometric or spatial network in a transport GIS is to model interconnected features, much like graphs in mathematics or computer science; important examples include the aforementioned road networks and utility networks (electricity, gas, water, etc.). Spatial networks have become essential to modern developments in fields such as energy, transport, and the environment. Indeed, the correct modeling of a geographical area of ​​interest makes it possible to solve countless technical problems such as: infrastructure planning, transport safety analysis, traffic monitoring and control, transport demand analysis, routing, and scheduling, and environmental impact assessment.

[0005] Nowadays, various spatial networks are being developed in the form of open or private data collections. Some well-known examples are OpenStreetMap™ (OSM), HERE Maps™, and Google Maps™. Unfortunately, to date, there is no standard network to which one can refer. This poses a crucial problem when it comes to integrating equivalent layers from different sources.

[0006] Today, metropolitan areas, consulting firms, and approved air quality monitoring associations (AASQA) rely on travel modeling and simulation tools to plan and simulate the impact of future traffic regulations and construction projects on congestion and air quality. These simulation tools require modeling a transport supply, in the form of a road or public transport network, onto which mobility demand is allocated. The road networks thus modeled are often derived from various cartographic sources and are enhanced by the aforementioned stakeholders with parameters specific to each study (e.g., traffic flow on network arcs, historical or measured average speed on network arcs, etc.).

[0007] These road networks, even when enriched with parameters specific to each stakeholder's expertise, are far from exhaustive. Indeed, it is often necessary to contextualize, further enrich, or improve the accuracy of the parameters of these road networks with other parameters from other data sources (e.g., signage, gradient, or number of lanes on the network arcs). The problem, therefore, lies in the lack of tools enabling direct correlation between the road arcs of a road network A with certain parameters and the road arcs of a road network B with other parameters and defined on a different cartographic reference system.

[0008] In order to be able to complete and improve (in particular the accuracy) the parameters available by merging several data sources, it is necessary to remove this obstacle related to the correspondence of networks defined on different cartographic references. Prior art

[0009] Mapmatching approaches found in the literature are limited to determining the correspondence between two road arcs solely based on their spatial proximity, sometimes adding their orientation as a discriminant. This leads to significant errors, especially for dense networks such as urban environments, where simple spatial proximity or orientation is insufficient to reduce uncertainty about the correspondence. In most applications, these mapmatching errors are not not critical. On the other hand, to determine an attribute of road traffic, matching a secondary road arc with a major axis due to their proximity alone would generate significant errors in the estimates of the quantities of pollutants emitted, for example, given the large difference in vehicular flow for the two arcs.

[0010] The integration, mapping, or alignment of different spatial networks representing the same geographic area is now considered a field in its own right. This is due to the importance of data transfer between networks and the impossibility of doing so manually for large datasets. The earliest recorded methods for network mapping probably date back to the remarkable work of Alan Saalfeld and his colleagues in the 1980s. These authors presented a semi-automatic method for mapping the features of two different maps in order to transfer information between them. The maps considered were the U.S. Geological Survey and the Washington, D.C., Census Bureau. The information transfer included road class, street names, and house number ranges.The techniques used include map alignment by "rubber-sheeting" (which can be translated as elastic transformation) and various elements from mathematical fields such as: topology, graph theory, statistics and optimization.

[0011] Modern matching methods tend to be separated in various ways, for example: deterministic and probabilistic, discrete and continuous, or geometric and topological. This complicates the correct classification of an ever-increasing body of literature on this subject. It is important to note, however, that similarity measures remain central to the network matching problem. They can be classified as geometric, topological, attribute, contextual, and semantic measures. Geometric measures are particularly relevant and are classified into several groups, including Euclid, Hausdorff, and Fréchet distances.Regarding matching, matching strategies tend to vary between one-to-one (1:1 - one strand to one strand), one-to-multiple (l:n - one strand to many strands), and many-to-multiple (n:m - many strands to many strands), depending on the problem specifications. Finally, matching methods can be evaluated using common statistical scores such as accuracy and recall.

[0012] The research problem initially motivated by the Census Bureau continues in different directions. The document "VOLKER W., DIETER F., Matching spatial data sets: a statistical approach, 1999, International Journal of Geographical Information Science, 13(5)" presents a matching approach functional for the integration of spatial data from different sources. The approach is based on statistical network studies, and the matching problem is transposed into a communication system where information theory tools are applied to find an optimal solution. The more abstract research developed in the paper: “ALT, H., EFRAT, A., ROTE, G., WENK, C., Matching planar maps, 2003, Journal of Algorithms, 49” provides powerful tools for measuring the similarity of line segment patterns in the plane. The distance measures used include generalizations of Fréchet distance and, interestingly, also consider the similarity between polygonal curves and graphs, or even between two graphs. The thesis: “ZHANG, M."Methods and Implementations of Road-Network Matching, 2009, PhD dissertation, Technical University of Munich" concerns an automatic matching approach for road networks. The author discusses the popularity of the "Buffer Growing" and "Iterative Closest Point" matching algorithms and their combinations. He proposes a new contextual approach (already commercialized) based on the "Delimited-Stroke-Oriented" algorithm to utilize valuable contextual information. A widely applied and successfully used matching method in the German region (TOMTOM, HERE maps, ATKIS) is studied in the document: "ZHANG, M., YAO, W., MENG, L., Automatic and Accurate Conflation of Different Road-Network Vector Data towards Multi-Modal Navigation, 2016, ISPRS Int. J. Geo-Inf., 5(5).” .

[0013] US patent application 2021231444 describes a method for similarizing road networks using distances based on geometry. This approach is based on different notions of distance that can be applied between geometric objects in road network representations. However, this method is not flexible enough in terms of applicability. In particular, this method requires significant computing resources (memory, processor) as well as a substantial computation time.

[0014] US patent 6564224 describes a method requiring fewer computing resources. This method breaks down the matching process by first considering the nodes of the road network. This approach allows for rapid matching despite the size of the road network. However, this speed comes at the expense of the method's robustness, due to the failure to account for complex geometries in the first step of the algorithm. Summary of the invention

[0015] The present invention relates to determining a road traffic attribute (e.g., vehicle flow, vehicle speed, quantity of pollutants emitted) from at least two representations of a transport network in a flexible, adaptable, robust manner, and with limited use of computing resources. To this end, the invention relates to a method for determining at least one road traffic attribute from at least two representations of the transport network, at least one road traffic parameter being associated with one of the two representations of the transport network. To determine the road traffic attribute, a third representation of the transport network is constructed by means of a correspondence between the two representations of the transport network, the correspondence being implemented after a strand elimination step (which uses at least one version of the discrete Fréchet distance).This strand elimination step simplifies the strand matching step by reducing the number of possible strands, thereby reducing the computing resources used and the computation time. Furthermore, this step ensures the robustness and adaptability of the process according to the invention.

[0016] The invention relates to a method for determining at least one road traffic attribute for a transport network, said transport network being represented by at least a first and a second transport network representation, each transport network representation comprising a plurality of strands, and at least one of said first or second transport network representations comprising at least one road traffic parameter associated with several strands of said transport network representation. For this method, the following steps are carried out: a. For each strand of said first transport network representation, strands of said second transport network representation are eliminated by means of at least one discrete Fréchet distance; b. For at least one strand of said first road network representation, a correspondence is determined between the strand of said first transport network representation and at least one strand of said second transport network representation among strands of said second transport network representation not eliminated; c. A third transport network representation is constructed using said first and second transport network representations and said determined correspondence between strands of said first and second transport network representations; and

[0017] The said road traffic attribute is determined on at least one strand of said third transport network representation by means of said third transport network representation and said road traffic parameter.

[0018] Advantageously, said road traffic attribute is chosen from a quantity of pollutants emitted by said road traffic, a flow of vehicles, a speed of vehicle traffic, a quantity of noise emissions emitted by said road traffic.

[0019] According to one embodiment, said road traffic parameter is chosen from a quantity of pollutants emitted by said road traffic, a vehicle flow rate, a vehicle traffic speed, a quantity of noise emissions emitted by said road traffic.

[0020] According to one implementation, said method includes a preliminary step of pre-processing said first and second transport network representations.

[0021] Advantageously, said pre-processing step includes a sub-step of cutting said first and second transport network representations into several transport network portions, and then strand elimination and strand matching steps are applied to each transport network portion.

[0022] According to one aspect, said first and second transport network representations include the names of said road strands, and the strand elimination step includes a substep of eliminating strands from said second transport network representation by comparing the names of the strands with those of said first transport network representation.

[0023] According to one embodiment, the strand elimination step further includes an elimination of strands from said second transport network representation based on a criterion of coverage of a strand of said first transport network representation by a strand of said second transport network representation.

[0024] According to one embodiment, the method further includes a step of evaluating the correspondence between said first and second transport network representations, before the step of constructing said third transport network representation, and according to said evaluation, the strand elimination and strand matching steps are repeated by reversing the order of said first and second transport network representations.

[0025] According to one implementation, said third transport network representation is constructed from a geometric representation of said first or said second transport network representation, or from a fusion of said first and second transport network representations.

[0026] According to one aspect, said method includes a preliminary step of acquiring said road traffic parameter.

[0027] According to one embodiment, said road traffic attribute is displayed on said third transport network representation by means of a computer system or a smartphone.

[0028] Other features and advantages of the process according to the invention will become apparent from the following description of non-limiting examples of embodiments, with reference to the figures attached and described below. List of figures

[0029] [Fig.1]

[0030] Fig. 1 illustrates the steps of the process according to a first embodiment of the invention.

[0031] [Fig.2]

[0032] Figure [Fig. 2] illustrates the steps of the process according to a second embodiment of the invention.

[0033] [Fig.3]

[0034] Figure 3 illustrates the steps of the process according to a third embodiment of the invention.

[0035] [Fig.4]

[0036] Fig. 4 illustrates, as a first example, a first representation of a road network.

[0037] [Fig.5]

[0038] Fig. 5 illustrates, for the first example, a second representation of a road network.

[0039] [Fig.6]

[0040] Fig. 6 illustrates, for the first example, five portions of the road network with the two representations of figures 4 and 5 superimposed.

[0041] [Fig.7]

[0042] Figure 7 illustrates, for the first example, the third network constructed using the method according to an embodiment of the invention.

[0043] [Fig. 8]

[0044] Figure 8 illustrates, for the first example, the average speed of vehicles on the road network determined by means of the method according to an embodiment of the invention.

[0045] [Fig.9]

[0046] Figure 9 illustrates, for the first example, the quantity of nitrogen oxides emitted by vehicles on the road network determined by means of the process according to an embodiment of the invention.

[0047] [Fig. 10]

[0048] Fig. 10 illustrates, for the first example, the quantity of carbon dioxide emitted by vehicles on the road network determined by means of the method according to an embodiment of the invention.

[0049] [Fig. 11]

[0050] Fig. 11 illustrates, for the first example, the quantity of unburned hydrocarbons emitted by vehicles on the road network determined by means of the process according to an embodiment of the invention.

[0051] [Fig. 12]

[0052] Fig. 12 illustrates, for a second example, a portion of a road network with superimposition of two road network representations.

[0053] [Fig. 13]

[0054] Fig. 13 illustrates, for the second example, a strand of the first representation to be matched with a strand of the second representation.

[0055] [Fig. 14]

[0056] Fig. 14 illustrates, for the second example, a strand of the second representation corresponding to the strand identified in Fig. 13 determined by means of the process according to an embodiment of the invention.

[0057] [Fig. 15]

[0058] Fig. 15 illustrates, for the second example, a strand of the second representation corresponding to the strand identified in Fig. 13 determined by means of a prior art process. Description of the implementation methods

[0059] The present invention relates to determining a road traffic attribute on a transport network, in particular a road network, a public transport network, or a soft mobility transport network (bicycle, scooter, etc.). A transport network is defined as a set of roads and paths in a predefined geographical area. A road network is a set of roads and paths in a geographical area that are passable by road vehicles. The term "road traffic" refers to the traffic of the mode of transport in question. This predefined geographical area may be a district of a city, a city, a group of municipalities, a department, etc. A strand of the transport network is an elementary subdivision of the transport network between two consecutive nodes of the transport network.For example, a segment of the transport network could be a road between two consecutive intersections, between two consecutive traffic signals, between an intersection and a traffic signal, or a section of highway between two consecutive exits, and so on. This provides a detailed breakdown of the transport network and a model that is suitable for the transport network without requiring microscopic data. In the following description, the road network is considered the transport network. However, the invention is applicable to other modes of transport such as public transport, active mobility, etc.

[0060] A road traffic attribute is a physical characteristic of road traffic on at least a section of the road network. This may include, in particular: - The quantity of pollutants emitted by vehicles on the road network, for example the quantity of nitrogen oxides, the quantity of carbon dioxide, the quantity of unburned hydrocarbons, etc. - Road traffic flow, that is to say the number of vehicles per unit of time that pass through a road network, - Vehicle speed, for example average speed or instantaneous speed, - The amount of noise emissions produced by vehicles on the road network, - A safety indicator (which can for example be defined by a rate linked to the level of congestion of the road segment, to the speed limit on the road segment, to the infrastructure of the road segment), etc.

[0061] The road network can be represented by a graph, called a road graph or road network representation. The road graph is composed of a set of edges and nodes, the nodes representing intersections, and the edges representing road segments (strands) between the intersections. The road graph can be obtained from an online mapping service (“web service”), for example Here TM (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 any two roads, and only those, are represented by the nodes of the graph), and as time-invariant as possible.Furthermore, the road graph can be simplified by not taking into account road sections such as dead ends, paths in parks or cycle paths, depending on the type of vehicle considered (for example, for the embodiment of motor vehicles, cycle paths may not be taken into account).

[0062] The present invention is implemented for at least two distinct road network graphs, referred to respectively as the first road network representation and the second road network representation. These two representations may originate from two different sources. Furthermore, at least one of the two road network representations includes at least one road traffic parameter. The two road network representations may include at least one road traffic parameter, which may be identical or different. The road traffic parameter is associated with a plurality of strands in the relevant road network representation. The road traffic parameter may be chosen, in particular, from among: - The quantity of pollutants emitted by vehicles on the road network, for example the quantity of nitrogen oxides, the quantity of carbon dioxide, the quantity of unburned hydrocarbons, etc. - Road traffic flow, that is to say the number of vehicles per unit of time that pass through a road network, - Vehicle speed, for example average speed or instantaneous speed, - The amount of noise emissions emitted by vehicles on the road network, etc.

[0063] According to one embodiment of the invention, the method may include a preliminary step of acquiring at least one road traffic parameter. Preferably, this step of acquiring at least one road parameter may include a measurement of the road traffic parameter at at least one point on the road network. Thus, the road traffic parameter can be accurate and reliable, and can allow for a precise determination of the road traffic attribute of the road network. Alternatively or cumulatively, this step of acquiring at least one road parameter may include acquisition from a geographic information system (GIS). Here™, Google Maps™, and OpenStreetMap™ are examples of geographic information systems. Road traffic parameters are usually available from any location.

[0064] According to a non-limiting embodiment, if the first road network representation includes vehicle speed as a road traffic parameter, if the second road network representation includes the speed limit, and if the road traffic attribute is vehicle speed, then the vehicle speed can be determined on the strands of the third road network representation, taking the speed limit into account. Thus, the vehicle speed can be more reliable because it depends on the speed limits, which was not the case in the first representation.

[0065] In addition, other macroscopic road network data can be associated with the road segments of at least one of the two road network representations: for example, topology data (i.e., gradient, road curvature, intersections, signage, etc.), the type of road segment (e.g., urban road, motorway, etc.), the number of lanes in the road segment, the maximum speed of the road segment, the signage of the road segment, the gradient of the road segment, the length of the road segment, the types of vehicles permitted, infrastructure present on the segment that could influence driving (bus stops, service stations, charging stations, signage, etc.), etc. These are essentially parameters related to the road network infrastructure. Preferably, the macroscopic road network data can be provided by a geographic information system (GIS).Here™, Google Maps™, and OpenStreetMap™ are examples of geographic information systems. Macroscopic data is usually available from any location.

[0066] The present invention therefore relates to a method for determining at least one Road traffic attribute on a road network. This process implements the following steps: 1. Removal of strands 2. Matching the strands 3. Construction of the third representation 4. Determining the attribute

[0067] These steps can be implemented by computer means, for example a computer or a server. These steps will be detailed later in the description.

[0068] Figure 1 illustrates, schematically and without limitation, the steps of the process for determining at least one road traffic attribute according to a first embodiment. A first representation of the road network (RES A) and a second representation of the road network (RES B) are implemented in this process. At least one of these two road graphs has at least one road traffic parameter. Then, for each strand of the first representation of the road network (RES A), strands of the second representation of the road network (RES B) are eliminated, notably by means of a version of the discrete Fréchet distance. A correspondence (COR) is then determined between strands of the first representation of the road network (RES A) and strands of the second representation of the road network (RES B) among the strands that were not eliminated.This correspondence allows us to construct a third road network representation (RES C), and to determine at least one road traffic attribute (ATT).

[0069] According to one embodiment of the invention, the method may further include a preliminary preprocessing step for the first and second road network representations. Thus, the shape of the two representations can be adapted to match the strands and at least one road traffic parameter. This preprocessing step may include a substep of dividing each road network representation into several road network segments. Then, the strand elimination and strand matching steps can be applied to each road network segment. This step makes it possible to limit the area over which the strand matching is determined, thereby reducing the computing resources required (memory and processors), as well as the computation time.

[0070] For this implementation, the method may include the following steps: A. Pre-processing of road network representations 1. Removal of strands 2. Matching the strands 3. Construction of the third representation 4. Determining the attribute

[0071] These steps can be implemented by computer means, for example a computer or a server. These steps will be detailed later in the description.

[0072] Figure 2 illustrates, schematically and without limitation, the steps of the process according to this implementation. Steps identical to those in Figure 1 are not detailed again. The process includes an additional pre-processing step for the road network representations, with a substep for cutting (DEC) the road network representations (RES A; RES B). Then, the steps for removing strands (ELI) and matching strands (COR) are applied to the cut portions of the road network.

[0073] According to one embodiment of the invention, the method may further include a step for evaluating the correspondence between the first and second road network representations. This evaluation step can be implemented between the step of matching the strands and the construction of the third road network representation. If the evaluation is satisfactory, the process can proceed to the step of constructing the third representation. If the evaluation is unsatisfactory, the steps of eliminating the strands and matching the strands can be repeated, reversing the order of the first and second road network representations. In this way, the robustness and reliability of the matching of the two road graphs can be improved, as well as the robustness and accuracy of the determination of the road traffic attribute.

[0074] For this implementation, the process may include the following steps: • 1. Removal of strands • 2. Matching the strands • B) Correspondence assessment • 3. Construction of the third representation • 4. Determining the attribute

[0075] These steps can be implemented by computer means, for example a computer or a server. These steps will be detailed later in the description.

[0076] Figure 3 schematically illustrates, without limitation, the steps of the process according to this embodiment. Figure 3 reproduces the embodiment of Figure 2; however, this embodiment is also adapted to the embodiment of Figure 1. The steps identical to those of Figure 2 are not detailed again. The process includes an additional step of evaluating (EVA) the correspondence of the strands (COR). If the evaluation is satisfactory, the process continues with the step of constructing the third road network representation (RES C). If the evaluation is unsatisfactory, the order of the road network representations (RES A) and (RES B) is reversed, and the steps of eliminating the strands (ELI) and matching the strands (COR) are repeated, as well as, where applicable, the substep of cutting (DEC) the road network representations.

[0077] A) Pre-processing of road network representations

[0078] In this optional step, the first and second road network representations are preprocessed to homogenize the two road graphs. This embodiment is particularly suitable for cases where the data concerning the road network representations tend to include different information and parameters in different formats. This step then allows the data to be prepared in the correct format.

[0079] According to one embodiment, each road segment can be described by a sequence of latitude and longitude pairs. Each sequence of latitude and longitude points must have a minimum of two coordinates, corresponding to the nodes of the network. The number of points available per road segment has a direct impact on the accuracy of the road network integration. Advantageously, each road segment can be described, if available, by the names of the corresponding roads (also called segment names). Providing road names can significantly improve the matching.

[0080] According to one aspect of the invention, this step may further include a map projection or the use of a geographic coordinate system to establish an identical geographic reference frame.

[0081] According to one implementation, this step may further include a substep for interpolating road segments. Indeed, the density of the point sequence defining a road segment can vary not only depending on the road network, but also depending on the specific segment. This preliminary substep aims to standardize the resolution of road segments by distance interpolation.

[0082] Furthermore, the preprocessing may further include the substep of dividing the road network representations into road network segments. This substep aims to reduce the computing resources and computation time of the invention by dividing the matching algorithm into different segments or zones. For example, the zones may be rectangular, square, circular, triangular, or of any similar shape. According to one embodiment, one may begin by selecting the dimensions of the road network segments surrounding each strand of the first road network representation. For example, one may define a road network segment having an envelope of 100 m (alternatively 50, 200, or 500 m, etc.) around a given road strand. Other methods may be implemented.Then, the same selection of road network segments is reproduced on the second road network representation, such that the road network segments of the second road network representation correspond to the road network segments of the first road network representation. 1. Removal of strands

[0083] In this step, for each strand of the first road network representation, strands from the second road network representation are eliminated using at least one discrete Fréchet distance (for example, according to a version of this distance as described below). In other words, in this step, for each strand of the first representation, the strands from the second representation that least match the considered strand of the first representation are removed, according to a criterion defined by the discrete Fréchet distance. Strand elimination is performed for each strand of the first road network representation. Thus, this step limits the number of strands to be compared in the strand matching step.

[0084] For the embodiment in which the substep of cutting into road network portions is implemented, this strand elimination step can only be implemented for strands belonging to the same road network portion.

[0085] According to one embodiment, this strand elimination step can implement a sequence of substeps, each substep being designed to eliminate strands. Thus, several elimination methods are applied, allowing for better strand selectivity, enabling better strand matching, and consequently, better accuracy of the road network attribute. This sequence of substeps can include at least one of the following substeps: - Elimination based on the discrete Fréchet distance, - Elimination based on the road strand name, - Elimination based on a coverage criterion, etc.

[0086] These substeps can be implemented in this order, or in any other order. Each substep compares the criterion to a predetermined threshold to eliminate strands from the second road network representation that exceed this predetermined threshold. For example, for the discrete Fréchet distance: strands whose distance exceeds a predetermined threshold can be eliminated.

[0087] Preferably, the strand removal step implements at least the substep based on the discrete Fréchet distance and the substep based on a coverage criterion. Thus, at least two removal methods are used, which increases the number of strands removed.

[0088] It is recalled that the Fréchet distance between paths A and B can be informally described as the length of the shortest leash allowing a person on path A to walk a dog on path B (without reversing). A discrete modified Fréchet distance is defined as a determination of a distance based on the method used to calculate the Fréchet distance, but not considering the greatest distance between the points on the paths of the two considered routes, but rather a sum of the distances between the two considered routes. in a discreet manner. The measurement of dynamic time distortion (DTW), introduced in the field of speech recognition as a measure of similarity between time series, is also relevant.

[0089] For the discrete Fréchet distance, we construct in the same way as for the classical Fréchet distance: - a polyline connecting the points of the strand to be characterized (first representation), - a polyline connecting the points of the strand (second representation), and then - the set of links that connect a point of each polyline.

[0090] The step of determining the links connecting a point on each polyline is performed dynamically and recursively. Two discrete strands X and Y of respective lengths N and M are considered here. The first step concerns the construction of a cumulative distance matrix, denoted D, which contains in row n, and column m, the discrete Fréchet distance calculated from the beginning of each path to points n and m (where 1 > n ≤ N and 1 / m < My). We can then write:

[0091] D^n, m) = F / Xn, Ym)

[0092] With F d the discrete version of the Fréchet distance, Xn, Ym the truncated paths of length n and m. The difference between the continuous and discrete versions of the Fréchet distance comes from the fact that we do not consider the parameterizations of the Fréchet distance as continuous functions of time but only the ordering of the links of the paths.

[0093] The matrix can be calculated recursively as follows:

[0094] D^d^y^+mint D^j, D^D^,

[0095] Where the indices i and j are between 1 and the length of each trajectory, and where x and y represent the ordered points on each trajectory. The initialization is defined as:

[0096] Mi.o — O, D| o = IXj — oo,

[0097] and the traditional similarity result is then simply given by the upper right corner of this matrix D.

[0098] Finally, to define a discrete Fréchet distance (modified in the context of this application), it can be made sensitive to subtrajectories. Indeed, it often happens that road segments of very different sizes represent the same street. To this end, the optimal coupling can be restricted by imposing a saturation on the number of links per point: each point can be coupled with at most a fixed number of points on the opposite trajectory. This prevents the last point of the smallest strand or sub-strand from being associated with all the other points of the largest strand, which would otherwise increase the trajectory distance. According to one implementation, the distance A modified discrete Fréchet function can be obtained by taking the mean, a percentile, and the median of the sequence of distances resulting from the optimal coupling. This further enhances the robustness of the similarity function.

[0099] According to one embodiment, the Haversian distance and / or the Euclidean distance or similar distance can be used to determine the Fréchet distance.

[0100] The substep of eliminating strands based on strand names (e.g., street names) is optional, as it can be implemented if the first and second road network representations include the names of the road strands. This similarity can be determined by a method for calculating an edit distance. For example, the Levensthein distance, which is a distance in the mathematical sense, provides a measure of the difference between two strings (in this case, the strings of the strand names). This distance is equal to the minimum number of characters that must be deleted, inserted, or replaced to go from one string to the other. As another example, the Wagner-Fischer algorithm, which is an algorithm for calculating the edit distance between two strings, can be applied.Any other editing distance or analogous algorithm can be implemented for this substep.

[0101] A substep for eliminating strands based on coverage can also be implemented. Coverage is defined as the portion (i.e., the length) of the considered strand from the first road network representation that is covered (overlapped) by a strand from the second road network representation. In other words, coverage can be defined as the ratio of the length of the strand from the first road network representation associated with a strand from the second road network representation to the total length of the strand from the first road network representation. Coverage can be defined as a percentage. This coverage can be determined when determining the discrete Fréchet distance, particularly under the version described above. Indeed, the version of the Fréchet distance presented can include a saturation on the number of links per point.Therefore, points on a strand of the first road network representation may not be linked to any point on a strand of the second road network representation (for example, if the strand in the first road network representation is shorter than the strand in the second road network representation). This substep allows for prioritizing longer strands during the matching step.

[0102] According to one embodiment, this elimination step can further take into account the connectivity between the strands. In other words, the first and second road network representations can include information relating to the strand connectivity (in other words, the strands entering and exiting each strand under consideration). In this case, the strand removal step can take this connectivity into account to remove strands, particularly strands that do not have the same strands entering and / or exiting the strand under consideration. 2. Matching the strands

[0103] In this step, for each strand of the first road network representation, a match is determined between the strand in question from the first road network representation and at least one strand from the second road network representation. This match is determined from among the strands not eliminated in step 1. In other words, among the strands not eliminated in the previous step, at least one strand from the second representation that most closely matches each strand of the first representation is determined. The strand matching is implemented for each strand of the first road network representation. This step can be implemented by optimizing at least one criterion used for strand elimination: for example, the minimum discrete Fréchet distance of the uneliminated strands, the maximum coverage rate of the uneliminated strands, and / or the minimum edit distance of the uneliminated strands.Thanks to the sorting performed during the strand elimination step, the number of possible strands is minimized, allowing this strand matching step to be carried out robustly and reliably, with limited computing resource requirements. It is possible that a strand in the first road network representation may not have corresponding strands in the second road network representation.

[0104] For the embodiment in which the substep of cutting into road network portions is implemented, this strand matching step can only be implemented for strands belonging to the same road network portion.

[0105] According to one embodiment, this matching step can further take into account the connectivity between the strands. In other words, the first and second road network representations can include information relating to the connectivity of the strands (that is, the strands entering and exiting each strand considered). In this case, the strand matching step can take this connectivity into account to match strands, in particular strands having the same strands entering and / or exiting the strand considered. B) Correspondence assessment

[0106] In this optional step, the strand matching determined in step 2 is evaluated. For example, this evaluation could take the form of determining the percentage of strands that have been matched for each road network representation. Indeed, the percentage of matched strands from one or the other of the representations The matchability of the road network may be low. This can be due to discontinuities in the strands or missing strands. If the evaluation is satisfactory (percentage of matched strands greater than a predetermined threshold), the process can continue with step 3. If the evaluation is unsatisfactory (percentage of matched strands less than a threshold), steps 1 and 2 of strand elimination and strand matching can be repeated, changing the order of the first and second road network representations. In other words, for the iteration of the steps, the second road network representation becomes the first road network representation, and vice versa. In this way, the robustness and reliability of the matching of the two road graphs can be improved, as well as the robustness and accuracy of the determination of the road traffic attribute.When the steps are repeated by modifying the order of the road network representations, the multi-matching of the two road network representations is also achieved. Identical matches, if any, can be deleted once the steps are completed.

[0107] In addition, a symmetrical version of the Fréchet distance orientation can also be applied if the direction of the road is not correctly described by the road network representations. 3. Construction of the third representation

[0108] During this step, a third road network representation is constructed, using the two road network representations and the strand correspondence determined in step 2. In other words, a third road network representation is constructed which corresponds to the correspondence of the two initial road network representations, so as to have the parameters of the two road network representations in a single geographical reference frame.

[0109] For the construction of the third road network representation, the strands can be constructed as: - Strands of the first representation of the road network, or - Strands of the second road network representation, or - Merging of the strands of the first and second road network representations, for example, this may be an average of the coordinates of the corresponding strands of the two road network representations (possibly after resampling the strands on an equal number of points), or by any similar method.

[0110] In addition, the third road network representation may include at least one macroscopic parameter related to one of the first and second road network representations (for example, topology). 4. Determining the attribute

[0111] In this step, at least one road traffic attribute is determined using the third representation constructed in step 3, and using at least one road traffic parameter from one of the road network representations. Preferably, the road traffic attribute can be determined for a plurality of strands of the third road network representation. Depending on the type of the at least one road traffic parameter and the type of the road traffic attribute, the road traffic attribute can correspond directly to the road traffic parameter and is then associated with the third road network representation, or it can be determined from this parameter.Thus, the method according to the invention makes it possible to obtain at least one road traffic attribute and a new augmented road graph which may not correspond to any of the input road graphs but which contains the information of all of them, in particular the road traffic parameters.

[0112] According to one implementation, in which the same type of road traffic parameter is associated with the strands of the first and second road network representations, the road traffic attribute can be determined from the road traffic parameters of each representation, for example by means of an average, possibly by means of a weighting, or any other similar method. By taking into account at least two road traffic parameters, the accuracy of the determination of the road traffic attribute is improved.

[0113] According to one aspect, for which a road traffic parameter is available on only one of the two initial road network representations, the attribute can be determined directly from the road parameter, and it is then associated with the third road network representation.

[0114] According to one embodiment, the determination of the road traffic attribute may further depend on at least one macroscopic parameter related to one of the first and second road network representations. Indeed, the road traffic attribute may depend on a macroscopic parameter. For example, the vehicle flow rate may depend on the number of lanes in the segment, or the quantity of pollutants emitted may depend on the slope of the segment, or the vehicle speed may depend on the speed limit of the segment.

[0115] According to one embodiment, if a strand of the third representation corresponds to at least two strands of the second representation, each having an associated road traffic parameter, then the road traffic attribute of the strand of the third representation can be determined from the parameters of each relevant strand of the second representation, for example by means of an average, possibly by means of a weighting, or any other similar method. By taking into account several road traffic parameters, the accuracy of the determination of the road traffic attribute

[0116] According to a non-limiting embodiment, if the first road network representation includes vehicle speed as a road traffic parameter, if the second road network representation includes the speed limit, and if the road traffic attribute is a quantity of pollutants emitted, then the vehicle speed can first be determined on the strands of the third road network representation, taking into account the speed limit (for example, by "capping" the speed at the speed limit), and then a model of pollutants emitted by the vehicles can be applied. Such a model links vehicle speed to a quantity of pollutants; it can be a macroscopic emissions model, such as the COPERT model described in the document: "Ntziachristos, L., Gkatzoflias, D., Kouridis, C., & Samaras, Z. (2009). COPERT: a European road transport emission inventory model."In Information technologies in environmental engineering (pp. 491-504) », or the model described in one of the patent applications FR3049653 (US2019138669) or FR3122011 (US2022335822). .

[0117] Thus, thanks to this step, the correspondence makes it possible to contextualize and explain mobility flows or polluting emissions using new road traffic attributes and therefore to extrapolate or robustify estimates to other areas having similar attributes.

[0118] For the embodiment in which the method is implemented for a number of transport network representations greater than two, the steps of the method can be repeated for matching the transport network representations two by two, until a single resulting representation is obtained.

[0119] Furthermore, the method according to the invention may include a step of displaying the road traffic attribute. During this step, the determined road traffic attribute is displayed on a transport map, in particular a road map (on a road graph corresponding to the third road network representation). This display may take the form of a rating, a color code, or a thickness of the road representation. This display may be carried out on board the vehicle: on the dashboard, on a portable, standalone device, such as a geolocation device (such as a GPS), or a mobile phone (such as a smartphone). It is also possible to display the road traffic attribute on a website. In addition, the road traffic attribute may be shared with public authorities (for example, road managers) and public works companies.Thus, public authorities and public works companies can determine which roads have a high road traffic attribute (e.g., flow, speed or quantity of pollutants emitted) and adapt the roads to users (e.g., creation of new lanes, modification of signage, etc.).

[0120] Furthermore, the invention relates to a method for managing the infrastructure of a road network. For this method, the following steps can be implemented: a. At least one road traffic attribute is determined by means of a road traffic attribute determination method according to any one of the variants or combinations of variants described above; and b. At least one infrastructure of the road network is modified according to the determined road traffic attribute, for example an infrastructure for which the flow of vehicles, the speed of vehicles or the quantity of pollutants emitted is maximum or greater than a predetermined threshold.

[0121] Thus, a road network can be managed to limit or even avoid pollution peaks, traffic jams, and the risk of accidents.

[0122] According to one embodiment, the modification of the infrastructure can be chosen in particular from the addition of signage (speed limit, light, yield, stop, etc.), the construction of a new lane, conversion of a strand to one direction, construction of a new road, etc. Examples

[0123] The characteristics and advantages of the method according to the invention will become clearer upon reading the examples below.

[0124] The first example concerns the correspondence of two complementary road network representations of the same location.

[0125] For this example, the pollutant emission model described in patent application FR3122011 (US2022335822) is applied to determine the different quantities of pollutants emitted for the segments of a road network. This model was developed so that characteristics such as gradient, speed limit, and average speed per segment (as well as additional information such as time of day, vehicle distribution, etc.) can predict these types of emissions. The road networks provided by the Here Technologies group incorporate this information by road segment or segment. However, it is known that the average speed estimates from this tool tend to be inaccurate, affecting the determination of the emission quantity provided by the macroscopic model. Therefore, it is necessary to obtain better speed measurements from another representation of the road network.

[0126] For this example, we consider the case of the neighborhoods around the "Croix-Rousse" district in the Lyon metropolitan area. There is a representation of this road network generated using BDTOPO® technology, in which precise average speed data has been acquired and associated with the segments of this representation. This representation is denoted BDTOPO. Thus, the BDTOPO network corresponding to this area The calculation includes actual speeds measured per strand according to the time period. Unfortunately, this additional information cannot be directly taken into account because the road networks of Here and BDTOPO have significant differences.

[0127] In summary, the objective is to maximize the accuracy of predicting pollutant emissions per strand on a road network representing an area around the "Croix-Rousse" district. For this example, the road network representation is designated network A. To achieve this objective, a good estimate of the average speed per strand is required. This necessary information is contained in a different road network representation (BDTOPO or network B) which represents the same geographical location. Network A, for its part, allows for the inclusion of other parameters necessary for the accurate determination of pollutant emissions, namely, road gradient and speed limits.

[0128] Fig. 4 illustrates, schematically and in a non-limiting manner, network A.

[0129] Fig. 5 illustrates, schematically and in a non-limiting manner, network B.

[0130] It can be noted that network A is larger than network B. Therefore, the The steps of the process according to the invention are implemented on the area defined by network B.

[0131] Figure 6 illustrates some portions of the road network with the superimposition of network A (RES A), represented by the black lines, and network B (RES B), represented by the grey lines. It can be observed that not all road segments of network B overlap with segments of network A. It is therefore necessary to correlate these two networks.

[0132] The method according to the invention is then applied, taking into account, for network B, the speed of the vehicles as a road traffic parameter. For this example, the names of the strands were not considered: the elimination of the strands was implemented using the Fréchet distance and coverage by means of predefined thresholds. Furthermore, since the desired degree of continuity was not achieved (evaluation step), the strand elimination and strand matching steps were repeated considering network B as the first road network representation, and considering network A as the second road network representation, i.e., by reversing the road network representations. The method makes it possible to construct the third road network representation, denoted network C, and then to determine several road traffic attributes: the speed of the vehicles as well as the quantities of pollutants emitted by the vehicles.For this example, we determine the road traffic attributes during the morning rush hour.

[0133] Fig. 7 illustrates, schematically and in a non-limiting manner, the network C thus constructed.

[0134] Figure 8 schematically illustrates, without limitation, the vehicle speed V on network C determined by the method according to the invention, based in particular on speed information from network B. For this example, vehicle speed is a road traffic attribute, which was imprecise on network A alone. The shade of gray of the road strands represents vehicle speed. The darkest strands have a speed V3 between 24 and 48 km / h, the intermediate gray strands have a speed V2 between 15 and 24 km / h, and the lightest strands have a speed VI between 3 and 15 km / h.

[0135] These speed indications come from network B, and serve as input to the pollutant emission model, to determine the amount of nitrogen oxide (NOx) emitted, the amount of carbon dioxide (CO2) emitted and the amount of unburned hydrocarbons (HC) emitted which are also road traffic attributes.

[0136] Figure 9 illustrates, schematically and without limitation, the quantity of nitrogen oxide emitted on network C determined by the method according to the invention from the average speed and the pollutant emission model. For this example, the quantity of nitrogen oxide is a road traffic attribute. The shade of gray of the road segments represents the quantity of NOx emitted. The darkest segments have an NOx quantity of between 954 and 1800 mg / km, the intermediate gray segments have an NOx quantity of between 637 and 954 mg / km, and the lightest segments have an NOx quantity of between 180 and 637 mg / km.

[0137] Figure 10 schematically illustrates, without limitation, the quantity of carbon dioxide emitted on network C determined by the method according to the invention based on the average speed and the pollutant emission model. For this example, the quantity of carbon dioxide is a road traffic attribute. The shade of gray of the road segments represents the quantity of CO2 emitted. The darkest segments have a CO2-3 quantity between 350 and 550 g / km, the intermediate gray segments have a CO2-2 quantity between 250 and 350 g / km, and the lightest segments have a CO2-1 quantity between 75 and 250 mg / km.

[0138] Figure 11 schematically illustrates, without limitation, the quantity of unburned hydrocarbons emitted on network C as determined by the method according to the invention, based on the average speed and the pollutant emission model. For this example, the quantity of unburned hydrocarbons is a road traffic attribute. The shade of gray of the road segments represents the quantity of HC emitted. The darkest segments have an HC-3 quantity between 79 and 133 mg / km, the intermediate gray segments have an HC-2 quantity between 63 and 79 mg / km, and the lightest segments have an HC-1 quantity between 36 and 63 mg / km.

[0139] Thus, this first example shows that the method according to the invention makes it possible to determine at least one attribute of the road network in a robust manner.

[0140] The second example is a comparative example that compares the method according to the invention with an alternative method for matching conventional road networks, namely the method used by Here™. For this example, the same networks A and B are used as in the first example. The drawbacks of this tool's method are numerous. First, the matching implemented in Here has a different objective and is therefore not optimized for the task of matching networks; the goal is to match GPS trajectories to their road network. This is why, for this example, the strands of a second network can be considered as GPS trajectories, thus resulting in a multi-matching (n:m) between networks. Second, and most importantly, it is not possible to match two arbitrary networks without first passing through the Here network, which can tend to drastically amplify their inaccuracies.This issue also exists for other road network matching tools such as "SharedStreets"™.

[0141] Here's matching method is based on a hidden Markov model that determines the path with the highest probability and provides a confidence index for each match. Low confidence indices can be observed in certain critical situations (intersections, junctions, exits, etc.) where the method according to the invention tends to succeed.

[0142] Figure 12 schematically and non-limitingly illustrates a portion of the road network around the Fourvière tunnel in Lyon. In this figure, it can be seen that networks A and B are not superimposed, which therefore necessitates aligning the road network representations.

[0143] Fig. 13 is an enlargement of Fig. 12, on which a BRT test strand is represented, for which the two methods of matching road network representations will be compared.

[0144] Fig. 14 illustrates the same area, and represents in discontinuous lines the BR-INV strand determined as corresponding by means of the method according to the invention.

[0145] Fig. 15 illustrates the same area, and represents in discontinuous lines two pieces of BR-AA strands determined using the classical Here method.

[0146] It can be observed from this comparison that the method according to the invention allows for a suitable match of the test strand, whereas the prior art method does not allow for such a match. Consequently, the method according to the invention is more reliable than the prior art methods.

Claims

Claims

1. Method for determining at least one road traffic attribute (ATT) for a transport network, said road traffic attribute being chosen from a quantity of pollutants emitted by said road traffic, a flow rate of vehicles, a speed of vehicle circulation, a quantity of noise emissions emitted by said road traffic, said transport network being represented by at least one first and one second distinct transport network representation (RES A, RES B), each transport network representation (RES A, RES B) comprising a plurality of nodes representing intersections of said transport network and strands representing road portions, and at least one of said first or second transport network representations (RES A, RES B) comprising at least one road traffic parameter associated with several strands of said transport network representation (RES A, RES B),said road traffic parameter being chosen from a quantity of pollutants emitted by said road traffic, a flow of vehicles, a speed of vehicle circulation, a quantity of noise emissions emitted by said road traffic, characterized in that the following steps are implemented by computer means:, a. The at least one road traffic parameter is acquired, including a measurement of the road traffic parameter at at least one point on the road network; b. For each strand of said first transport network representation (RES A, RES B), strands of said second transport network representation are eliminated (ELI) by means of at least one discrete Fréchet distance, such that strands of the second representation (RES B) which correspond least to the strand considered of the first representation (RES A) are eliminated; c. For at least one strand of said first representation (RES A) of road network, a correspondence (COR) is determined between the strand of said first representation (RES A) of transport network and at least one strand of said second representation (RES B) of transport network among strands of said second representation (RES B) of transport network not eliminated, by determining at least one strand of the second representation which is closest to said strand of said first representation; d. A third representation (RES C) of transport network is constructed by means of said first and second representations (RES A, RES B) of transport network and said correspondence (COR) of said strands determined between strands of said first and second representations (RES A, RES B) of transport network; and e. Said road traffic attribute is determined on at least one strand of said third representation (RES C) of transport network by means of said third representation (RES C) of transport network and said acquired road traffic parameter.

2. Method according to one of the preceding claims, wherein said method comprises a prior step of pre-processing said first and second transport network representations.

3. Method according to claim 2, wherein said preprocessing step comprises a sub-step of cutting (DEC) said first and second representations (RES A, RES B) of transport networks into several transport network portions, and then applying the steps of eliminating (ELI) strands and matching (COR) strands for each transport network portion.

4. Method according to one of the preceding claims, wherein said first and second representations (RES A, RES B) of transport network comprise the name of said road strands, and the step of eliminating (ELI) strands comprises a sub-step of eliminating strands from said second representation (RES B) of transport network by comparing the names of the strands with those of said first representation (RES A) of transport network.

5. Method according to one of the preceding claims, in which the step of eliminating (ELI) strands further comprises eliminating strands from said second transport network representation (RES B) according to a criterion of coverage of a strand of said first transport network representation (RES A) by a strand of said second transport network representation (RES B).

6. Method according to one of the preceding claims, in which the method further comprises a step of evaluating (EVA) the corres- correspondence between said first and second transport network representations (RES A, RES B), before the step of constructing said third transport network representation (RES C), and depending on said evaluation (EVA), the steps of strand elimination and strand matching are repeated by reversing (INV) the order of said first and second transport network representations.

7. Method according to one of the preceding claims, in which said third representation (RES C) of transport network is constructed from a geometric representation of said first or said second representation (RES A, RES B) of transport network, or from a fusion of said first and second representations (RES A, RES B) of transport network.

8. Method according to one of the preceding claims, in which said road traffic attribute is displayed on said third transport network representation by means of a computer system or a smartphone.

Images