METHOD FOR DETERMINING AT LEAST ONE ROUTE BETWEEN A GEOGRAPHIC AREA OF ORIGIN AND A GEOGRAPHIC DESTINATION AREA

Description

technical field

[0001] The present invention relates to the field of determining a route between a geographical area of ​​origin and a geographical area of ​​destination.

[0002] The main area of ​​application is mobility analysis in terms of choosing the most probable routes for travel between two areas of a road network. This could involve, for example, determining travel patterns between a residential area and a business park. Such mobility analysis allows for ex-post and ex-ante assessments of the impact of public policies regarding planning and regulation on mobility, as well as on pollutant emissions, air quality, and traffic forecasting.

[0003] Indeed, the choice of routes for the population becomes a key factor for the analysis of transport networks and their performance, for example, in terms of total travel time, total distance travelled and / or energy consumption.

[0004] This analysis of travel between two areas also allows for a behavioral assessment of users with regard to the choice of route for their daily travel. Previous technique

[0005] Current air quality monitoring tools cannot accurately isolate and estimate the proportion of pollutant emissions from road transport during actual use, nor their spatial location. Energy and emissions diagnostic tools for mobility (EDEM) exist that can currently estimate the impact of an average trip between two areas of a city, but this is done by assuming a straight-line distance and a constant average speed throughout the journey. If the two origin-destination areas are large and / or distant, this state-of-the-art method necessarily neglects all the variability in terms of possible routes between the two areas, as well as the contribution of each route to the total emissions of inter-zone trips.

[0006] Consequently, it proves difficult for cities to make the right decisions regarding road infrastructure planning and legislation without having accurate tools for assessing and projecting the impact of planned measures on road transport pollution emissions and air quality.

[0007] The study of mobility patterns or route choice is a recurring topic in transport science, and it can be divided into two categories.

[0008] The first category relies on so-called discrete choice models, which aim to explain individual choice behavior through individual preferences and a utility function. This approach to route choice analysis tends to be based on limited possibilities, considering a small number of routes and origin / destination points, as exemplified in the document BERGMAN CECILIA ET AL: "Conflation of OpenStreetMap and Mobile Sports Tracking Data for Automatic Bicycle Routing", TRANSACTIONS IN GIS, vol. 20, no. 6, March 10, 2016 (2016-03-10), pages 848-868. The second category aims to solve the so-called network load problem; it proposes a more comprehensive approach to determine the static or dynamic states of traffic across the entire transport network and, ultimately, the network's equilibrium.This is made possible by considering demand in relation to route choice among all origin / destination pairs in the set of zones considered. The drawback of this second approach is that the implemented mathematical representation, based on deterministic or stochastic users, tends not to take into account all relevant variables due to simplification arguments.

[0009] Furthermore, data technology has experienced exponential growth in recent years; we are now able not only to record very large datasets concerning very different aspects of human mobility, but also to analyze them effectively using artificial intelligence or machine learning. Mobility data can be collected in various forms, and its characteristics can vary enormously. Collection is often carried out using cell phones or embedded geolocation devices (e.g., GPS, or Global Positioning System). Common datasets include Global Positioning System (GPS) logs and Call Detail Records (CDRs). In general, this type of representation of mobility remains limited and imperfect.For example, GPS data with a frequency of 1Hz representing a user's movement may have widely inaccurate measurements (aberrations) or measurements absent for several seconds.

[0010] An unsupervised learning method for detecting mobility patterns relies on grouping travel into families that share spatiotemporal characteristics. Geographic trajectory grouping is a widely used technique for extracting valuable information from large datasets that are otherwise difficult to process. The first core task of the grouping approach is defining a function that efficiently determines trajectory similarities. Typical complications arise from the presence of noise in the data, outliers, and the fact that trajectories are usually formed by a different number of points.Popular choices for discrete trajectory similarity functions include: Euclidean distance, dynamic time warp, longest common subsequence, edit distance, Hausdorff distance, and Fréchet distance. The second key task of this approach is to develop the actual clustering algorithm, once the similarity function is properly defined. Generic clustering algorithms may be unsuitable for this task; for example, the K-means method is not robust to outliers and cannot produce convex-shaped clusters.A better option for trajectory clustering is the use of density-based clustering algorithms, especially those that reject noise, such as DBSCAN (density-based spatial clustering of applications with noise: it is a density-based algorithm insofar as it relies on the estimated density of clusters to perform the partitioning).

[0011] Measuring the distance between two trajectories can be computationally expensive. This can be particularly problematic when pooling huge datasets of trajectories, which are necessary for extracting useful mobility models from a population. In fact, most unsupervised pooling algorithms have a complexity greater than or equal to O ( N2< ) where N is the number of trajectories; this follows from the calculation of the N × N distance matrix formed by all pairwise distances. Traditionally, this limits studies to subsets of large datasets, since analyses of standard-sized trajectory datasets can take several days.

[0012] Even though it is possible to gather relatively large datasets of trajectories and group them effectively, a known challenge remains the representativeness and potential bias of the acquired data. Indeed, true mobility patterns can only be determined if the data themselves conform to reality.

[0013] US patent application 2013191314 relates to a method for determining a user's preferred route. This approach involves searching for the most probable travel route for a user between two geographic areas and then analyzing the characteristics associated with that route. This approach is limited to a single user and is not suitable for handling large amounts of data from multiple users to represent inter-area mobility on a large scale. Furthermore, generalizing from a single user to city-wide mobility is subject to representativeness issues or biases. US patent application 20060242108 relates to a data fusion method for route optimization, specifically for managing emergency-related movements.This approach is based on the use of various data sources, including a geographic information system (GIS), to study an optimal route for a given geographic area and event. However, the goal of this approach is not to analyze route choices, but rather to optimize variables associated with possible routes. Furthermore, data fusion is applied at the sensor level to increase data reliability. Therefore, this approach requires specific equipment, such as expensive ground sensors or cameras. Summary of the invention

[0014] The present invention aims to determine a route between two geographical areas in a representative and rapid manner, using limited computer memory and without requiring specific sensors placed on the road network. The invention relates to a method for determining at least one route between a geographical origin area and a geographical destination area. The method acquires measured geolocation data and aggregates this geolocation data. Furthermore, the method determines possible routes between the two areas and aggregates the data from these possible routes. Combining the representative trajectories resulting from the aggregated data (travels taken and routes) allows the routes between the two geographical areas to be deduced.Thus, the process implements a fusion of trajectories representing measured geolocation data and trajectories representing routes, ensuring robustness and representativeness of the determined route without requiring specific sensors on the road network. Furthermore, the grouping method employed eliminates the need for a map-matching step, thereby reducing computation time and the amount of computer memory used.

[0015] The invention relates to a method for determining at least one route between a geographical area of ​​origin and a geographical area of ​​destination according to claim 1.

[0016] According to one embodiment, said geolocation data of completed journeys and / or said geolocation data of determined routes are grouped together using a "Quick Bundles" type grouping algorithm.

[0017] According to the invention, said at least one route is deduced by means of a conflation method of said trajectories representing said journeys carried out and of said trajectories representing said itineraries.

[0018] Advantageously, prior to the grouping stage of said measured geolocation data of journeys made, said measured geolocation data of journeys made are segmented to retain relevant measured geolocation data, in particular by cutting out the measured geolocation data within said geographical area of ​​origin and / or within said geographical area of ​​destination.

[0019] According to one aspect, to determine at least two routes between said geographical area of ​​origin and said geographical area of ​​destination, a plurality of origin points are determined in said geographical area of ​​origin and a plurality of destination points in said geographical area of ​​destination from the origin and destination points of the trajectories representing said routes made.

[0020] Advantageously, prior to the stage of grouping determined routes, said determined routes are segmented to retain relevant portions of routes, in particular by cutting said routes within said geographical area of ​​origin and / or within said geographical area of ​​destination.

[0021] According to one implementation option, for each representative trajectory and / or for each road, at least one characteristic is determined, chosen from average speed, journey time, traffic.

[0022] Preferably, the steps are implemented for a given time period, such as a month, a week, a day, or a time slot.

[0023] According to one embodiment, after the grouping step of said measured geolocation data of journeys made, only the groupings for which the number of journeys made and / or the number of users of said journeys made is greater than or equal to a predetermined threshold are retained.

[0024] Furthermore, the invention relates to a method for determining the quantity of at least one chemical and / or noise pollutant emitted by at least one vehicle during a journey within a road network between a geographical area of ​​origin and a geographical area of ​​destination, according to claim 12, as well as a method for determining the energy consumption of at least one vehicle within a road network during a journey between the geographical area of ​​origin and the geographical area of ​​destination, according to claim 13.

[0025] Other features and advantages of the process according to the invention will become apparent from the following description of non-limiting examples of implementations, with reference to the figures attached and described below. List of figures

[0026] There figure 1 illustrates the steps of the process according to a first embodiment of the invention. figure 2illustrates the steps of the process according to a second embodiment of the invention. figure 3 illustrates the steps of the process according to a third embodiment of the invention. figure 4 illustrates a trajectory map representative of paths taken for the application of the process according to one embodiment of the invention, for example. figure 5 illustrates a map of trajectories representing routes for applying the process according to an embodiment of the invention, for example, the figure 4 . There figure 6 illustrates a map with the routes obtained by applying the method according to an embodiment of the invention, for example, Figures 4 and 5 . There figure 7 illustrates a map with the routes obtained by applying a process according to a prior art, for example, figures 4 to 6 . Description of the implementation methods

[0027] The invention relates to a method for determining at least one route for a vehicle between a geographical origin area and a geographical destination area. A geographical area is defined as a given territory, for example, a region, a department, a group of municipalities, a city, a district of a city, etc. A geographical area therefore comprises a plurality of points (respectively, a plurality of origin points and a plurality of destination points for journeys). By way of example only, this could involve determining journeys between a residential geographical area and a geographical area of ​​professional activities. The geographical origin area corresponds to the territory from which a journey begins, and the geographical destination area corresponds to the territory to which a journey ends.Thus, the invention makes it possible to determine route choices that users of a road network could make between any pair of points originating from the geographical origin-destination zones. The geographical origin and destination zones can be polygonal in shape. The geographical origin and destination zones can be adjacent or distant.

[0028] Preferably, the vehicle is a motorized vehicle circulating within the road network, such as a car, a two-wheeler, a heavy goods vehicle, a coach, or a bus. Alternatively, the vehicle may be a soft mobility vehicle: a bicycle, a scooter, etc.

[0029] According to the invention, the following steps are implemented: 1. Acquisition of measured geolocation data 2. Grouping of measured geolocation data 3. Determination of trajectories representative of the measured geolocation data 4. Determination of routes 5. Grouping of routes 6. Determination of trajectories representative of the routes 7. Determination of roads

[0030] These steps can be implemented using computer technology, specifically a computer and / or a server. These steps are detailed later in the description.

[0031] In this application, the term "journey" defines the movements made and measured, the term "route" defines movements obtained by a geographic information system, the term "trajectory" defines a movement representative of a grouping, and the term "route" defines a movement obtained by the method according to the invention.

[0032] There figure 1This illustrates, schematically and without limitation, the steps of the process according to a first embodiment of the invention. From the origin and destination geographic areas ZON, measured geolocation data MES is acquired. The measured geolocation data MES is grouped REG1. For each group of measured geolocation data REG1, a representative trajectory TRA1 is determined. From the origin and destination geographic areas ZON, at least two routes SIG are determined. The routes are grouped REG2. For each group of routes, a representative trajectory TRA2 is determined. Then, the representative trajectories obtained from the measured geolocation data and the routes are merged FUS to determine at least one route rte between the origin geographic area and the destination geographic area.

[0033] There figure 2illustrates, schematically and without limitation, the steps of the process according to a second embodiment of the invention. The steps are identical to the embodiment of the figure 1 are not described again. For this second embodiment, the GIS determination of at least two routes takes into account the representative trajectories of the journeys undertaken. In particular, the representative trajectories (TRA) can be used to define origin and destination points for the GIS determination of the routes.

[0034] In one embodiment, a segmentation step may be included between the geolocation data acquisition step and the measured data grouping step, and / or between the route determination step and the route grouping step. For this embodiment, the process may include the following steps: 1. Acquisition of measured geolocation data 1'. Segmentation of measured geolocation data 2. Grouping of measured geolocation data 3. Determination of trajectories representative of the measured geolocation data 4. Determination of routes 4'. Segmentation of routes 5. Grouping of routes 6. Determination of trajectories representative of routes 7. Determination of roads

[0035] These steps can be implemented using computer technology, specifically a computer and / or a server. These steps are detailed later in the description.

[0036] There figure 3 illustrates, schematically and without limitation, the steps of the process according to a third embodiment of the invention. The steps are identical to the embodiment of the figure 2are not described again. The process includes a segmentation step (SEG1) of the measured geolocation data (MES), followed by a grouping step (REG1) that considers the segmented geolocation data. The process also includes a segmentation step (SEG2) of the GIS routes, followed by a grouping step (REG2) that considers the segmented routes.

[0037] In accordance with an implementation plan, the steps of the process can be carried out for a given time period. For example, for a month, a week, a day, or a time slot. This implementation allows for consideration of the variability of traffic conditions (school holidays, public holidays, seasons, months, peak hours, etc.). 1. Acquisition of measured geolocation data

[0038] During this step, geolocation data is acquired from journeys taken between the origin and destination geographic areas. This journey data may include data measured during previous journeys, such as speed, position, and altitude. Preferably, the journey data can be measured using a geolocation system, such as a satellite positioning sensor like GPS (Global Positioning System), Galileo, etc. The geolocation system can be integrated into the vehicle or remote (for example, via a smartphone).

[0039] According to one embodiment, during this step, geolocation measurements can be taken during the movement of at least one vehicle.

[0040] Preferably, this geolocation data can be in the form of time-stamped sequences of latitude / longitude pairs (paths), so that values ​​such as duration and average speed can be estimated for each sequence. To maximize the amount of available data, paths passing through the selected areas can also be considered, even if their starting and ending points are not located within those areas.

[0041] Advantageously, the geolocation data can be unprocessed, meaning it is not mapped to an underlying transport network and / or contains outliers. In other words, in this embodiment, the process does not include a map-matching step for the measured geolocation data, thus reducing computation time and the amount of computer memory used.

[0042] For the embodiment in which the steps are implemented for a given time period, during this step, one can acquire the measured geolocation data of the journeys made which correspond to the given time period. 1'. Segmentation of measured geolocation data

[0043] In this optional step, the measured geolocation data of completed journeys is segmented to retain relevant geolocation data. This step allows for an appropriate segmentation of completed journeys to ignore irrelevant differences in the starting and ending points of the completed journeys.

[0044] In one embodiment, geolocation data segmentation can be implemented by dividing the measured geolocation data within the geographic origin area and / or within the geographic destination area. This segmentation can preferably be implemented for geographic origin and destination areas that are not adjacent. In other words, this segmentation allows for the preservation of portions of the journey taken between the geographic origin and destination areas.

[0045] According to one embodiment, for both neighboring and non-neighboring zones, an alternative partitioning can be performed that allows the data to remain within the geographical zones for the portion shared between the completed journeys. To do this, the two points of convergence between the journeys (the geographical zone of origin and the geographical zone of destination) are found by filtering out the outliers in the sequence d* as defined for the Fréchet distance below. 2. Grouping of measured geolocation data

[0046] In this step, the measured geolocation data of the completed (and, where applicable, segmented) routes, acquired in step 1, are grouped into representative clusters of completed routes. In other words, this step forms clusters of similar completed routes. This step can be implemented by determining the similarity between the completed routes. This similarity can be determined using the Fréchet distance between the completed routes. Thus, the method for identifying similar routes preferably uses the Fréchet distance. The Fréchet distance is a widely used symmetrical measure of the similarity between two curves that depends only on the location and order of the points. It is generally illustrated as the distance of the shortest leash that allows a person to walk their dog along their separate paths.Alternatively, the process can implement other distances, such as Euclidean distance, dynamic time warp, longest common subsequence, editing distance, Hausdorff distance.

[0047] In its continuous form, the Fréchet distance is given by the formula: F A B = inf α , β max i ∈ 0 1 d A α t , B β t where A and B are curves on a metric space (distance d), and where α and β represent different parameterizations of curves.

[0048] For example, the Fréchet distance can be constructed using the following steps: We form a polyline connecting the points of a first path, we form a polyline connecting the points of a second path, and then we form the set of links that connect a point of each polyline.

[0049] The step of determining the connections linking a point on each polyline is performed dynamically and recursively. Two discrete paths X and Y of respective lengths N and M are considered here. The first step involves constructing a cumulative distance matrix, denoted D, which contains, in row n and column m, the discrete Fréchet distance calculated from the start of each path to points n and m (where 1 ≤ n ≤ N and 1 ≤ m ≤ M). We can then write: D n m = F d Xn Ym

[0050] With F d The discrete version of the Fréchet distance, Xn, Ym, are the truncated paths of length n and m. The difference between the continuous and discrete versions of the Fréchet distance stems from the fact that parameterizations are not considered. α And β like continuous functions of time but only the ordering of the links of the paths.

[0051] The matrix can be calculated recursively as follows: D n m = min D n − 1 , m − 1 , D n − 1 , m , D n , m − 1 + d x n y m

[0052] Or xn corresponds to the nth element of the path X and ym the mth element of path Y, and d the distance between the nth element of path X and the mth element of path Y.

[0053] The construction of the links is then carried out in the reverse direction (starting from points N and M to the beginning of the two paths) by defining the pairs of points with the minimum distance, i.e.: P i − 1 = arg min D n − 1 , m − 1 , D n − 1 , m , D n , m − 1

[0054] We then obtain the optimal sequencing P = [ P n , ... P 1] where each Pi represents a pair of points xk and yl. In parallel, it is possible to obtain the optimal distance sequence d * = [ dn, ... , d 1] where di represents the metric distance between points xk and yl associated with PiFinally, the discrete Fréchet distance is obtained by taking the average of the distance sequence d*.

[0055] In one embodiment of this step, the measured geolocation data of the traveled paths can be grouped using an unsupervised learning method, such as a clustering algorithm, specifically a Quick Bundles algorithm. This Quick Bundles algorithm is a recursive algorithm based on the distance between streamlines and an associated similarity threshold. The centroids of the lines (i.e., the lines representing a cluster formed by the algorithm) are generated on the fly and compared to the incoming flow lines until they are all grouped. In this embodiment, the Quick Bundles algorithm can be applied to the latitude / longitude coordinate pairs of the measured geolocation data.The Quick Bundles algorithm allows for the processing of a large amount of geolocation data, with fast computation time and limited computer memory.

[0056] The paper "E. Garyfallidis, M. Brett, M. Correia, G. Williams, I. Nimmo-Smith, "QuickBundles, a Method for Tractography Simplification", Frontiers in Neuroscience, 2012, vol. 6, no. 175" presents the generally linear time-domain algorithm called QuickBundles (QB) for the massive clustering of neuroanatomical bundles represented by diffusion MRI datasets. The algorithm's time performance is further improved by a factor of 20 in the paper "E. Garyfallidis, M.-A. Côté, F. Rheault, M. Descoteaux, "QuickBundlesX: Sequential clustering of millions of streamlines in multiple levels of detail at record execution time", 24th International Society of Magnetic Resonance in Medicine (ISMRM), 2016".

[0057] In one respect, at the end of this step, only the groups for which the number of trips made within the group and / or the number of different users within the group is greater than or equal to a predetermined threshold can be retained. Thus, the most representative groups are kept. 3. Determination of trajectories representative of the measured geolocation data

[0058] In this step, for each representative grouping (for each cluster) determined in step 2, a representative trajectory is defined. In other words, for each cluster, a trajectory is defined that connects the geographical area of ​​origin to the geographical area of ​​destination and that represents the cluster.

[0059] Advantageously, the representative trajectory can be derived from the grouping algorithm. For the embodiment implementing the Quick Bundles algorithm, the representative trajectory can correspond to the centroid obtained by this algorithm. For each grouping, the centroids can be calculated as follows: 1) resampling the trajectories so that they have the same number of points, 2) averaging the latitude and longitude per numbered data point.

[0060] In accordance with the implementation of this step, at least one characteristic of the representative trajectory of the cluster can be determined, as well as, potentially, a distribution of this characteristic. This characteristic can be chosen from among the average speed, the travel time of the trajectory, the length of the trajectory, traffic, or the utilization rate (also called the proportion of use) of this representative trajectory, etc. The average speed and travel time of the trajectory can be determined, in particular, from the time data linked to the measured geolocation data of the completed journeys belonging to the group under consideration. The length of the journey can also be determined from the measured geolocation data belonging to the group under consideration.The traffic and utilization rate of this route can be determined, in particular, from the number of geolocation data points measured within the cluster. In other words, the larger the cluster (i.e., the more routes it contains), the higher the traffic and utilization rate. 4. Route planning

[0061] During this step, at least two routes are determined between the origin and destination geographic areas using a geographic information system (GIS). Here Maps, Google Maps, and OpenStreetMap are examples of geographic information systems. Thus, during this step, the geographic information system determines at least two possible paths between the origin and destination geographic areas. The geographic information system may, in particular, implement a shortest path algorithm, an algorithm that minimizes vehicle fuel consumption, an algorithm that minimizes the amount of pollutants emitted, etc. Therefore, the invention takes into account both measured data and data obtained through the GIS.In this way, the process makes it possible to determine a route, while reducing uncertainty and increasing reliability by merging two sources of information: a GIS route calculation and geolocation data measured from real use in a territory.

[0062] The origin and destination points of the routes belong respectively to the geographical areas of origin and destination.

[0063] According to one embodiment of the invention, the origin and destination points of the routes can correspond to the origin and destination points of the representative trajectories determined in step 3 for the routes taken. Thus, the determined routes start and end at the same place as the representative trajectories.

[0064] Alternatively or cumulatively, the origin and destination points of the routes may be arranged respectively near the border of the geographical areas of origin and destination, in particular when these geographical areas are not adjacent.

[0065] Alternatively, or cumulatively, the origin and destination points can be the vertices of the polygons in the origin and destination geographic areas. This arrangement eliminates the need for travel within the origin and destination geographic areas.

[0066] For the embodiment in which the process is implemented for a given time period, this route determination can be repeated for different times within that time period. For example, if the time period considered is a morning, routes can be generated for different times spaced thirty minutes apart. 4'. Route segmentation

[0067] In this optional step, the routes determined in step 4 are segmented to retain relevant portions of the routes. This step allows for an appropriate segmentation of the routes to ignore irrelevant differences in the starting and ending points.

[0068] In one embodiment, route segmentation can be implemented by dividing routes within the geographical origin area and / or within the geographical destination area. This segmentation can preferably be implemented for geographical origin and destination areas that are not adjacent. In other words, this segmentation allows for the preservation of route segments between the geographical origin and destination areas.

[0069] According to one embodiment, for neighboring or non-neighboring zones, an alternative partitioning can also be performed that preserves the routes. To do this, the two points of convergence between the routes (geographic origin zone and geographical destination zone) are found by filtering out the outliers in the sequence d* as defined for the Fréchet distance above. 5. Grouping of routes

[0070] In this step, the routes (and, where applicable, segmented routes) determined in step 4 are grouped into representative clusters. In other words, this step forms "clusters" (groupings) of routes. This step can be implemented by determining the similarity between the completed routes. This similarity can be determined using the Fréchet distance between the completed routes. Thus, the method for identifying similar routes preferably uses the Fréchet distance. Alternatively, the process can implement other distances, such as the Euclidean distance, dynamic time warp, longest common subsequence, edit distance, or Hausdorff distance.

[0071] According to one embodiment of this step, routes can be grouped using an unsupervised learning method, such as a clustering algorithm. 6. Determining trajectories representative of the routes

[0072] In this step, for each representative grouping (for each cluster) determined in step 5, a representative trajectory is defined. In other words, for each cluster, a trajectory is defined that connects the geographical area of ​​origin to the geographical area of ​​destination and that represents the cluster.

[0073] Advantageously, the representative trajectory can be derived from the grouping algorithm. For the embodiment implementing the Quick Bundles algorithm, the representative trajectory can correspond to the centroid obtained by this algorithm. For grouping, the centroids can be calculated as follows: 1) resampling the trajectories so that they have the same number of points, 2) averaging the latitude and longitude per numbered data point.

[0074] In accordance with the implementation of this step, at least one characteristic of the representative trajectory of the cluster can be determined, as well as, potentially, a distribution of this characteristic. This characteristic can be chosen from among the average speed, the travel time of the trajectory, the length of the trajectory, the traffic, or the utilization rate (also called the proportion of use) of this representative trajectory. The average speed and travel time of the trajectory can be determined, in particular, from the time data of the routes belonging to the group under consideration. The length of the route can be determined, in particular, from the routes belonging to the group under consideration. The traffic and the utilization rate of this trajectory can be determined, in particular, from the number of routes sampled within the period under consideration.In other words, the larger the cluster size (i.e., the more routes it contains at different time intervals), the higher the traffic and utilization rate. 7. Route Determination

[0075] In this step, at least one route, preferably a plurality of routes, is determined that connects the geographical area of ​​origin and the geographical area of ​​destination. This is done by comparing and combining the representative trajectories of the completed journeys determined in step 3 with the representative trajectories of the routes determined in step 6. Thus, at the end of this step, one or more routes are determined from among the representative trajectories identified in the previous steps.

[0076] According to one embodiment of the invention, a characteristic and optionally the distribution of this characteristic for each determined route can be determined during this step. This characteristic can be chosen, in particular, from among the average speed, the travel time of the trajectory, the length of the trajectory, the traffic, or the utilization rate (also called the proportion of use) of this route.

[0077] In this step, we can match the centroids (i.e., the representative trajectories) of each collection (one collection corresponds to the completed journeys and the other to the routes) by minimizing the Fréchet distance and considering a distance threshold. In this way, most of the centroids of the collections are matched; however, some centroids may remain unmatched and suggest new paths for the other collection. In other words, there could be more representative trajectories for the routes than for the completed journeys, and vice versa.

[0078] Next, the representative trajectories are merged to determine at least one route. This merging can also be applied to the feature determined in step 3 or 6. According to the invention, a merging method is implemented using a "conflation" method. Such a method is described, in particular, in the document: "TP Hill, J. Miller, "How to combine independent data sets for the same quantity", Chaos: An Interdisciplinary Journal of Nonlinear Science, 2011, vol. 21, no. 3, p. 033102." The conflation method makes it possible to combine different probability density functions representing a value while avoiding the usual drawbacks of other methods such as averaging probabilities or data.Conflation can be defined as the process of combining geographic information from overlapping sources to maintain accurate data, minimize redundancy, and resolve data conflicts. Thus, conflation is an optimal consolidation technique for distributions, summarizing multiple input data distributions into a single posterior distribution. This technique differs from standard statistical methods based on point estimates or confidence intervals. Unlike prior art methods (point or confidence interval estimation), this method does not introduce unnatural transformations of variance, increase uncertainty, or alter the overall shape of the distribution. Among the important properties of this method, the merging of normal distributions is again normal.

[0079] The statistical fusion of information between GIS routes and real-world geolocation data can, in particular, allow for the generation of probability distributions for travel time and average speed for each of the routes identified by the method. Finally, the invention can be used to define a proportion among the identified routes, that is, a "weight" for each route that depends on the number of users and journeys observed in the real-world data, as well as route suggestions from the GIS geographic information system used.

[0080] According to one embodiment of the invention, this fusion technique can also be applied locally, that is, by considering the fusion of information between sub-trajectories and the corresponding route links in the GIS. Finally, the proportion of travel flow between areas can be estimated by considering both the distribution of the actual data and the route information in the GIS.

[0081] According to one implementation of the invention, the proportion for a given time period TP and an indexed representative route RP can be calculated as follows. First, NP is defined as the total number of routes corresponding to TP, i.e., the number of centroids (representative paths) of completed route data plus the number of unidentified or new representative paths suggested by the GIS. If the representative route RP is a new path suggested by the GIS, then, for example, a uniform proportion of 100 / NP can be considered. If, on the other hand, RP is a representative centroid constructed from the measured geolocation data of actual routes, then the assigned proportion is given by a constant factor multiplying the ratio between the total number of representative paths corresponding to RP and the total number of paths corresponding to all centroids considered in the period TP.The constant factor is such that the sum of the proportions of all representative trajectories equals 100. This allocation of proportions takes into account both the actual data flow and its potential non-representativeness. However, instead of assuming a uniform distribution across the new routes, other types of information from the Geographic Information System (GIS) can potentially be used to determine more precise proportions, such as travel times, the difference between travel times on the routes, length, etc.

[0082] The process may include an optional step of displaying at least one specific road. During this optional step, the road(s) can be displayed on a road map. This display can take the form of a rating or a color code. If applicable, a rating or color can be associated with a characteristic (speed, traffic, etc.) of the road in question. This display can be implemented in a vehicle: on the dashboard, on a portable, standalone device such as a GPS tracking device, or a mobile phone (such as a smartphone). It is also possible to display the speed profile on a website. Furthermore, the specific road(s) can be shared with public authorities (e.g., road managers) and public works companies.Thus, public authorities and public works companies can optimize road infrastructure to improve safety or reduce polluting emissions.

[0083] The present invention also relates to a method for determining the quantity of at least one chemical and / or noise pollutant emitted by at least one vehicle within a road network during travel between the geographical area of ​​origin and the geographical area of ​​destination. This method for determining the quantity of at least one emitted pollutant involves the following steps: a) At least one route is determined between the geographical area of ​​origin and the geographical area of ​​destination by means of the method of determining at least one route according to one of the preceding characteristics, and a vehicle speed characteristic for the route is deduced from this; and b) a microscopic model of pollutant emissions, chemical and / or noise, is applied to the vehicle speed characteristic for the route to predict a quantity of at least one chemical and / or noise pollutant emitted on the road in question, the pollutant emission model being a model that links the vehicle speed to a quantity of chemical and / or noise pollutants emitted by the vehicle.

[0084] These steps can be implemented using computer technology.

[0085] For example, the microscopic model may correspond to the model as described in patent application FR3049653 (WO 2017 / 174239) or to the COPERT model (from the English "COmputer Program to calculate Emissions from Road Transports", which can be translated as computer program for calculating road transport emissions).

[0086] The application of the microscopic model can also take into account traffic and / or utilization rates, which can be obtained through route determination. Thus, the determined quantity of chemical and / or noise pollutant emissions is representative of all travel between the two geographical areas.

[0087] Advantageously, the process may include an optional step of displaying the quantities of chemical and / or noise pollutants emitted by at least one vehicle during a journey between the geographical area of ​​origin and the geographical area of ​​destination. During this optional step, the quantities of pollutants emitted can be displayed on a road map. This display can take the form of a rating or a color code. If necessary, a rating or color can be associated with each link in the road network. This display can be implemented on board a vehicle: on the dashboard, on a portable, stand-alone device, such as a GPS tracking device, or a mobile phone. It is also possible to display the pollutant emissions on a website.Furthermore, predicted pollutant emissions can be shared with public authorities (e.g., road managers) and public works companies. This allows them to optimize road infrastructure and reduce pollutant emissions.

[0088] The present invention also relates to a method for determining the energy consumption (e.g., fuel or electrical energy) of at least one vehicle within a road network during travel between the geographical area of ​​origin and the geographical area of ​​destination. This method for determining energy consumption involves the following steps: a) At least one route is determined between the geographical area of ​​origin and the geographical area of ​​destination by means of the method of determining at least one route according to one of the preceding characteristics, and a vehicle speed characteristic for the route is deduced; and b) a microscopic energy consumption model is applied to the vehicle speed characteristic for the route to predict energy consumption on the route in question, the energy consumption model being a model that relates vehicle speed to vehicle energy consumption.

[0089] These steps can be implemented using computer technology.

[0090] The application of the microscopic model can also take into account traffic and / or utilization rates, which can be obtained through route determination. Thus, the determined amount of energy consumption is representative of all travel between the two geographical areas.

[0091] Advantageously, the process may include an optional step of displaying the energy consumption of at least one vehicle for a journey between the geographical area of ​​origin and the geographical area of ​​destination. During this optional step, energy consumption can be displayed on a road map. This display can take the form of a rating or a color code. If necessary, a rating or color can be associated with each link in the road network. This display can be implemented on board a vehicle: on the dashboard, on a portable, standalone device such as a GPS tracking device, or a mobile phone (such as a smartphone). It is also possible to display the pollutant emissions on a website. Furthermore, the predicted pollutant emissions can be shared with public authorities (e.g., road authorities) and public works companies.Thus, public authorities and public works companies can optimize road infrastructure to improve polluting emissions. Examples

[0092] The advantages of the process according to the invention will become apparent from the application examples described below.

[0093] To illustrate the route determination method according to the invention, we consider a case of travel by motor vehicle between two given areas of the Lyon metropolitan area, as well as a chosen monthly time period and time slot on weekdays. For this example, the geographical origin area corresponds to the first and second arrondissements of Lyon, and the geographical destination area corresponds to the city of Bron.

[0094] Geolocation data for routes measured between January 1, 2017, and January 1, 2022, is acquired. This geolocation data consists of time-stamped GPS data. The defined time slots are from 6:00 AM to 12:00 PM and from 2:00 PM to 8:00 PM. Geolocation data outside these time slots is grouped together. The Geographic Information System (GIS) is the navigation tool derived from the HERE™ Maps application.

[0095] Since the two geographic areas are not adjacent, geolocation data is segmented to analyze mobility between areas rather than within them. Furthermore, path resampling is set to a default value of 256 points, and the distance threshold is 100 meters. It is important to note that a group of paths for a given time window is considered representative if there are at least 10 recorded paths taken by at least 5 different users.

[0096] The total number of trips completed, corresponding to the time period and the origin / destination geographic areas, is 2289. Applying the grouping step using the Quick Bundles algorithm then forms 424 clusters. Among these clusters, four are retained that meet the criterion stated above (at least 10 recorded trips completed by at least 5 different users). Then, the representative routes of each of these 4 clusters are determined.

[0097] There figure 4This schematic and non-exhaustive illustration shows a map of the Lyon metropolitan area with representative trajectories for each of the four retained clusters. To the west of the map is the origin zone (ZGO), which includes several origin points (PO). To the east of the map is the destination zone (ZGD), which includes several destination points (PD). The four representative trajectories for the retained clusters are labeled C1, C2, C3, and C4. Note that two trajectories, C1 and C2, pass through the south, while two trajectories, C3 and C4, pass through the north.

[0098] Next, a plurality of routes are determined using the HERE Maps application. To determine these routes, origin and destination points are considered that correspond to the origin and destination points of trajectories C1 to C4, as well as points on the boundary of the origin and destination geographic areas. Routes are determined for every 30 minutes within the chosen time slot. By applying the Quick Bundles algorithm, 12 clusters are obtained.

[0099] There figure 5 illustrates, schematically and in a non-exhaustive manner, the same map as the figure 4of the Lyon metropolitan area with representative trajectories of certain clusters (only 9 trajectories are illustrated for better readability of the figure). These representative trajectories are labeled I1 to I9. Note that representative trajectories I1, I2, I5 pass through the south, representative trajectories I3, I4, I9 pass through the north, and representative trajectories I6, I7 and I8 pass through the center.

[0100] By applying the conflation method, we merge the representative trajectories obtained from the measured geolocation data and the routes, and we determine a plurality of routes between the geographical area of ​​origin and the geographical area of ​​destination.

[0101] There figure 6 illustrates, schematically and in a non-exhaustive manner, the same map as the Figures 4 and 5with the routes determined by the method according to the invention. On this map, we find the representative trajectories C1 to C4 of the figure 4 and the trajectories I1 to I9 of the figure 5 We observe that the representative trajectories C1 to C4 are matched respectively with the representative trajectories I1 to I4, and that the process made it possible to identify other routes, notably route I5 (the southernmost), and routes I6, I7, I8 (in the center).

[0102] Thus, the method according to the invention makes it possible to identify more routes than with measured geolocation data alone. Furthermore, thanks in particular to the conflation method, it is possible to precisely determine characteristics related to travel on these routes, such as speed, traffic, etc.

[0103] For comparison, the HERE application proposes to determine routes between two geographical areas, by determining the origin point of the route as the barycenter of the origin geographical area, and the destination point of the route as the barycenter of the destination geographical area.

[0104] For this example of earlier art, the figure 7 illustrates, schematically and in a non-exhaustive manner, the same map as the figures 4 to 6 with the routes determined by the HERE application. These routes are indicated as T1 to T7. It is noted that the routes are different from the routes obtained by the method according to the invention. Moreover, these routes are different from the paths taken ( figure 4 Therefore, these routes are not representative of user movements.

[0105] The process according to the invention therefore allows for better representation than the prior art.

Claims

1. Method for determining at least one course (rte) between a geographical area of origin (ZGO) and a geographical area of destination (ZGD), characterized in that the following steps are implemented: a. measured geolocation data (MES) of at least two journeys completed between said geographical area of origin (ZGO) and said geographical area of destination (ZGD) are acquired; b. said geolocation data of completed journeys are clustered (REG1) into a plurality of clusters representative of journeys by means of a similarity between said completed journeys; c. for each cluster representative of journeys, a representative path (TRA1) is determined; d. at least two routes between said geographical area of origin (ZGO) and said geographical area of destination (ZGD) are determined by means of a geographic information system (SIG); e. said determined routes are clustered (REG2) into a plurality of clusters representative of routes of journeys by means of a similarity between said determined routes; f. for each cluster representative of routes, a representative path (TRA2) is determined; and g. said paths representative of said completed journeys and said paths representative of said routes are combined by fusion (FUS), and at least one course (rte) between said geographical area of origin (ZGO) and said geographical area of destination (ZGD) is deduced therefrom by means of a conflation method.

2. Method for determining at least one course according to Claim 1, wherein said geolocation data on completed journeys and / or said geolocation data on determined routes are clustered (REG1, REG2) by means of a QuickBundles clustering algorithm.

3. Method for determining at least one course according to either of the preceding claims, wherein, prior to the step of clustering (REG1) said measured geolocation data on completed journeys, said measured geolocation data on completed journeys are segmented (SEG1) so as to retain measured geolocation data relevant to geographical areas of origin and destination that are not neighbouring.

4. Method for determining at least one course according to any of the preceding claims, wherein to determine at least two routes between said geographical area of origin (ZGO) and said geographical area of destination (ZGD), a plurality of points of origin (PO) in said geographical area of origin (ZGO) and a plurality of points of destination (PD) in said geographical area of destination (ZGD) are determined from the points of origin (PO) and points of destination (PD) of the paths representative of said completed journeys.

5. Method for determining at least one course according to any of the preceding claims, wherein, prior to the step of clustering (REG2) determined routes, said determined routes are segmented (SEG2) so as to retain route segments relevant to geographical areas of origin and destination that are not neighbouring.

6. Method for determining at least one course according to Claim 3 or according to Claim 5, wherein the measured geolocation data on completed journeys and / or the determined routes are segmented (SEG1, SEG2) so as to retain relevant measured geolocation data and / or to retain relevant route segments, by breaking down said measured geolocation data and / or said determined routes within said geographical area of origin and / or within said geographical area of destination.

7. Method for determining at least one course according to any of the preceding claims, wherein at least one characteristic selected from average speed, journey time, journey length, traffic, and usage level, is determined for each representative journey and / or for each course, and wherein the association by fusion of step g) is also applied to said characteristic.

8. Method for determining at least one course according to any of the preceding claims, wherein the steps are implemented for a given time period.

9. Method for determining at least one course according to any of the preceding claims, wherein the steps are implemented for a month, or a week, or a day, or a time slot.

10. Method for determining at least one course according to any of the preceding claims, wherein after the step of clustering (REG1) said measured geolocation data on completed journeys, only clusters for which the number of completed journeys and / or the number of users of said completed journeys is greater than or equal to a predetermined threshold are kept.

11. Method for determining at least one course according to any of the preceding claims, wherein said similarity of said geolocation data on completed journeys and / or said similarity of said routes is / are determined by means of a Fréchet distance.

12. Method for determining an amount of at least one chemical pollutant and / or type of noise pollution generated by at least one vehicle during a trip within a road network between a geographical area of origin and a geographical area of destination, characterized in that the following steps are implemented: a. at least one course (rte) between said geographical area of origin and said geographical area of destination is determined by means of the method for determining at least one course according to any of the preceding claims, and a characteristic of the speed of the at least one vehicle for said at least one course is deduced therefrom; and b. a microscopic model of generation of chemical pollutants and / or types of noise pollution is applied to said at least one course so as to determine an amount of at least one chemical pollutant and / or type of noise pollution, said microscopic model relating the speed of the vehicle and the amount of at least one chemical pollutant and / or type of noise pollution generated by said at least one vehicle.

13. Method for determining the energy consumption of at least one vehicle within a road network when during a trip between the geographical area of origin and the geographical area of destination, characterized in that the following steps are implemented: a) at least one course between the geographical area of origin and the geographical area of destination is determined by means of the method for determining at least one course according to any of Claims 1 to 11, and a characteristic of the speed of the vehicle for the course is deduced therefrom; and b') a microscopic model of energy consumption at the speed characteristic of the vehicle for the course is applied to predict an energy consumption over the course in question, the model of energy consumption being a model that relates the speed of the vehicle to an energy consumption of the vehicle.