Method and system for assessing accident risk situations for a vehicle in circulation on a road network in a specific environment

Abstract

The invention relates to a method (50) and a system (10) for evaluating accident risk situations of a vehicle traveling on a road network in a determined environment, the method comprising: monitoring (51) the evolution of the vehicle on the road network and in the environment in which the road network is integrated and the vehicle moves, detecting (52) a risk situation among a plurality of accident risk factors (13); warning (53) the driver when a risk situation is recognized; analyzing (54) the recorded data which precede the detection (52) of the risk situation in order to determine an origin of the risk situation;and the production (55) of driving advice based on the determined contextual parameters (16), the driving advice identifies the detected risk situation and recommends driving behavior in order to reduce the risks incurred by the vehicle when approaching said situation. The invention also relates to a vehicle carrying the system (10) and executing the method (50). Figure for the abstract: Fig.1];

Description

Title of the invention: Method and system for evaluating accident risk situations for a vehicle traveling on a road network in a specific environment Technical field

[0001] The invention relates to a method and a system for evaluating accident risk situations for a vehicle in circulation on a road network in a given environment. The invention also relates to a land vehicle, in particular a motor vehicle, which incorporates a system and / or implements a method according to the invention. Prior art

[0002] Methods and systems are known for monitoring and evaluating the type of driving and the state of fatigue of the driver. These systems are based on capture members coupled to the actuators of the vehicle such as the accelerator pedal, the brake pedal. These systems also include optical sensors for detecting the blinking speed of the driver's eyes in order to detect in particular a state of drowsiness. These monitoring systems also include summary monitoring of the environment in which the vehicle is located, in particular, by monitoring the speed limits of the traffic lane on which the vehicle is moving but also the control of distances with a second vehicle which directly precedes or follows the vehicle in which the system is installed.These systems can also detect obstacles on the road using a distance sensor located at the front of the vehicle or the crossing of road markings using a camera. These systems monitor the vehicle's parameters to determine the driver's behavior (pressure on the accelerator pedal, steering wheel rotation speed, fuel consumption). However, these systems do not seek to improve the driver's behavior when approaching a risky situation such as an intersection recognized as an accident-prone area, or in the event of repeated traffic violations.

[0003] The invention aims to overcome all of these drawbacks. Statement of the invention

[0004] The invention aims to evaluate and improve the driver's behavior when approaching a specific risky situation.

[0005] The invention aims to provide a technical solution capable of providing an analysis of a risky situation encountered by the driver.

[0006] The invention aims to help the driver to improve his driving attitudes to approaching a risky situation.

[0007] In this regard, the invention relates to a method for evaluating accident risk situations for a vehicle in circulation on a road network in a given environment, the method comprises: - monitoring the vehicle's progress on the road network and in the environment in which the road network is integrated and the vehicle moves, monitoring is carried out by recording: vehicle movement data, in particular the vehicle's speed and its geolocation, road profile data that the vehicle is crossing, road signage data that the vehicle is crossing, data on the specifics of the environment of the road network in which the vehicle is moving, fixed or mobile obstacles that are located on the road network near the vehicle; - the detection of a risk situation among a plurality of accident risk factors including: traffic violations, risks of collision with a fixed or moving obstacle, risks of overspeeding, risks of hidden areas in the environment that the vehicle is crossing, risks of changing lanes, and risks of head-on collision on the same traffic lane, a risk situation being detected by comparing, in real time, the recorded data with the plurality of risk factors; - warning the driver when a risky situation is recognized; - the analysis of the recorded data which precede the detection of the risk situation in order to determine the origin of the risk situation, the analysis being carried out at least by determining contextual parameters of the risk: • a configuration of the route crossed using road profile data, • vehicle behavior when approaching said road configuration in using vehicle movement data, • a regulatory context linked to the configuration of the road crossed, the context being determined by signaling data, • the environment that the vehicle is crossing, data on the specific features of the road network environment make it possible to determine whether the vehicle is crossing an area in which certain risk factors are more or less likely, • the presence of obstacles likely to interact with the vehicle, the analysis of obstacle data makes it possible to define whether the alert is due to an interaction with fixed or mobile obstacles such as other road users; and - the production of driving advice based on the determined contextual parameters, the driving advice identifies the detected risk situation and recommends driving behavior in order to reduce the risks incurred by the vehicle when approaching said situation.

[0008] The method according to the invention makes it possible, through the detection of risk factors and the determination of the contextual parameters which define the risky situation, to evaluate the conditions in which the driver approached a risky situation, and thus to produce advice adapted to each situation. The adapted advice allows the driver to improve his driving attitude when approaching the same risky situation or a similar situation.

[0009] In embodiments, the production of the advice can be carried out from the risk situation by being defined by contextual parameters which have been determined and from an advice database which comprises: - predefined recommended driving behaviors when approaching the determined risk situation, and - the reasons why the said behaviors are recommended.

[0010] The method can thus construct suitable advice by selecting the recommended driving behavior and the most relevant reason based on the detected risk situation.

[0011] In embodiments, the production of the advice can be carried out by comparing said determined contextual parameters, with recommended behaviors and reasons stored in the database, the recommended behaviors and said reasons which correspond to said contextual parameters are associated with the detected risk situation in order to produce advice.

[0012] In embodiments, the evaluation method may comprise adjusting the relevance of the driving advice based on the driver's reaction when the warning is triggered for a given risky situation or when the driver provides feedback on the relevance of the driving advice after the method has questioned the driver. Adjusting the relevance P makes it possible to improve the efficiency of the method by personalizing the advice offered to the driver based on his feedback.

[0013] In embodiments, the evaluation method may comprise adapting the emission of warnings and the production of advice based on the behavior of the driver when approaching a risky situation that has already been detected in the past. The warnings may be modulated in intensity or frequency, this makes it possible not to over-solicit the driver for situations that present little risk or situations for which the driver improves his behavior when approaching them.

[0014] In embodiments, the recording of said data relating to the movement of the vehicle and its environment can be carried out using a buffer memory of a determined buffer duration which precedes the detection of the situation at risk, the buffer duration can be a few seconds, preferably ten seconds, so that when a risk situation is detected, the method analyzes the data recorded in the buffer. Using a buffer allows the analysis of the data recorded during the ten seconds preceding the detection of the risk situation and the issuance of a warning. It is thus possible to analyze the contextual parameters that led to the risk situation.

[0015] In embodiments, said data may comprise road profile data that is extracted from the recorded images of the road and map data based on the geolocation of the vehicle, the recorded images, the geolocation and the map data being correlated to determine the contextual parameter of the configuration of the road. The cross-referencing of data from several sources makes it possible to increase the reliability of the determination of the contextual parameter.

[0016] In embodiments, said data may comprise vehicle maneuver data including speed, braking, steering wheel angle, and turn signal data, the maneuver data being correlated to determine the contextual parameter of vehicle behavior when approaching said road configuration. The maneuver data makes it possible to define more precisely the driver's driving attitude when approaching the risky situation.

[0017] In embodiments, said data may comprise digital signaling data that are extracted from the map data based on the geolocation of the vehicle and visual signaling data that are extracted from the images recorded by the front camera of the vehicle, the data extracted from the images and the map data are correlated to determine the regulatory context related to the configuration of the road crossed. The signaling data extracted from the images may be analyzed by visual recognition algorithms that are known to those skilled in the art. Visual recognition algorithms are also known as "computer vision algorithms".

[0018] In embodiments, said data may comprise specificity data of the environment that the vehicle is crossing, the specificity data is extracted from cartographic data according to the geolocation of the vehicle, the specificity data comprises information on the types of buildings near roads, regulated speed zones, pedestrian crossings, the specificity data makes it possible to determine particularities of the environment that the vehicle is crossing which are linked to the regulatory contextual parameter. For example, it is possible to determine whether the vehicle is crossing an area in which certain risk factors are more or less probable such as an area recognized as ac- incidentogenic.

[0019] In embodiments, the contextual parameters may comprise temporal parameters which are determined by temporal data recorded during the tracking step, the temporal data providing information on the time and day at which the risk situation was detected, the temporal data being, in particular, correlated with the specificity data of the environment. For example, the method according to the invention can thus define increased pedestrian traffic if one passes near a station or a school at times during which the surroundings of these establishments are actively frequented.

[0020] In embodiments, said data may comprise fixed or mobile obstacle data that are located on the road network in the vicinity of the vehicle, the obstacle data are extracted from the recorded images, volumetric images of the road in front of the vehicle recorded by at least one volumetric detection sensor oriented in the direction of travel of the vehicle and distance data recorded by a distance sensor, the data extracted from the images, volumetric images and distance are correlated to determine the obstacle parameters, in particular, likely to interact with the vehicle in front of the vehicle. This data makes it possible to define whether there are obstacles on the road, their nature such as other vehicles, and their behavior relative to the behavior of the vehicle executing the method.

[0021] In embodiments, the contextual parameters may comprise parameters of position and direction of travel of the vehicle on lanes of the road, these parameters are determined by traffic data comprising a position of the vehicle on lanes of the road, and a direction of travel of the vehicle on this lane, the traffic data are extracted from the geolocation, cartographic data and images recorded by the camera, this traffic data makes it possible to define whether the vehicle has crossed a marking line on the ground and / or whether the vehicle was traveling in the correct direction of travel.

[0022] In embodiments, the contextual parameters may include weather parameters that are determined by weather data recorded during the tracking step, the weather data is extracted based on the vehicle's geolocation from a weather data website. Weather data can have a significant impact on traffic conditions and how a risky situation should be approached. By evaluating these parameters, the method makes it possible to refine the assessment of the situation and produce advice that is better suited to the situation.

[0023] In embodiments, the plurality of accident risk factors may include: - risks of collision with a fixed or moving obstacle which are detected when the distance to an obstacle reduces in a given time, - risks of overspeeding which are detected when the vehicle exceeds the speed limits or when the vehicle is moving fast in relation to the environment in which the vehicle is moving, - risks of hidden areas in the environment that the vehicle crosses, the hidden areas are determined by comparing the map data with the geolocation of the vehicle and the recorded images, - lane change risks, lane change is determined by comparing map data based on the vehicle's geolocation and recorded images, and - risks of frontal collision on the same traffic lane which are determined by recorded images and distance sensors located at the front of the vehicle and map data.

[0024] In embodiments, the collision risk factor may be weighted by a severity factor making it possible to quantify the seriousness of a possible collision, this severity factor may use as weighting criterion vehicle movement data such as speed, and / or data relating to the nature of the obstacle with which the risk of collision is detected.

[0025] In embodiments, the warning may include visual stimuli emitted in particular by a light strip, vehicle control equipment equipped with a haptic function such as the steering wheel, the accelerator pedal, the brake pedal, drawings and icons stored in a memory and displayed on a screen integrated into a passenger compartment of the vehicle, and sounds emitted by loudspeakers also integrated into the passenger compartment of the vehicle.

[0026] The invention also relates to a land vehicle, in particular an automobile, characterized in that it implements the method for evaluating accident risk situations for a vehicle in circulation on a road network in a determined environment which is defined according to the invention.

[0027] The invention also relates to a system for evaluating situations involving an accident risk for a vehicle in circulation on a road network in a given environment, the system comprising: - devices for capturing the vehicle's movement on the road network and in the environment in which the road network is integrated, the capture devices being configured to record vehicle movement data, in particular the vehicle's speed and its geolocation, road profile data that the vehicle is crossing, signaling data of the roads crossed by the vehicle, data on the specific environment of the road network in which the vehicle is moving, fixed or mobile obstacles located on the road network near the vehicle; - A control unit which is configured to detect a risk situation among a plurality of accident risk factors including: traffic violations, risks of collision with a fixed or moving obstacle, risks of overspeeding, risks of hidden areas in the environment that the vehicle is crossing, risks of changing lanes, and risks of head-on collision on the same traffic lane, the control unit being configured to detect a risk situation by comparing, in real time, the recorded data with the plurality of risk factors; - driver warning means which are configured to warn the driver of the imminence of a risky situation, - a module for analyzing the recorded data preceding the detection of the risk situation in order to determine the origin of the risk situation, the analysis module being configured to determine at least contextual parameters of the risk: • a configuration of the route crossed using road profile data, • vehicle behavior when approaching said road configuration using vehicle movement data, • a regulatory context linked to the configuration of the road crossed, the context being determined from the signaling data, • the environment that the vehicle is crossing, data on the specific features of the road network environment make it possible to determine whether the vehicle is crossing an area in which certain risk factors are more or less likely, • the presence of obstacles likely to interact with the vehicle, the analysis of obstacle data to determine whether the alert is due to an interaction with fixed or moving obstacles such as other road users; and - an advice module configured to produce driving advice based on the determined contextual parameters, the driving advice identifies the detected risky situation and recommends driving behavior in order to reduce the risks incurred by the vehicle when approaching said situation.

[0028] The system is configured to implement the method which is the subject of the invention, it has mirror characteristics compared to the characteristics of the method. In fact, these features address the same technical problem and present the same advantages. Brief description of the drawings

[0029] Other characteristics and advantages of the invention will become apparent from reading the description which follows. This description is purely illustrative and should be read in conjunction with the appended drawings in which:

[0030] [Fig. 1] is a representation of a flowchart illustrating the method and system for evaluating accident risk situations for a vehicle in circulation according to the invention.

[0031] [Fig.2] is a schematic representation of the risk factors and contextual parameters that are taken into consideration in producing advice to the driver in accordance with one embodiment of the invention.

[0032] [Fig.3] is a schematic representation of the risk factors and contextual parameters that are taken into consideration in producing advice to the driver in accordance with another embodiment of the invention.

[0033] [Fig.4] is a schematic representation of the structure of advice produced by the method and system for assessing accident risk situations in accordance with one embodiment of the invention.

[0034] [Fig.5] is a schematic representation, according to a flowchart, of the method for evaluating situations at risk of accident according to an embodiment of the invention. Description of the embodiments

[0035] With reference to [Fig.l], the invention relates to a system 10 for evaluating situations involving an accident risk for a vehicle traveling on a road network in a given environment. The invention also relates to a method 50 for evaluating situations involving an accident risk for a vehicle traveling on a road network in a given environment. Furthermore, the invention relates, of course, to a vehicle which incorporates the system 10 and implements the method 50 which is the subject of the present invention.

[0036] As illustrated in [Fig.l], the system 10 comprises elements 11 for capturing the vehicle's movement on the road network and in the environment in which the road network is integrated. In particular, the capturing elements are configured to record vehicle movement data such as the vehicle speed, steering wheel angle, braking, turn signals, windshield wipers, dipped headlights, etc. In general, the aforementioned data are available on the vehicle's electronic communication system; they are known as the "Data Bus". Those skilled in the art know how to retrieve them in real time using a specific algorithm.

[0037] The capture devices may also comprise means for geolocating the vehicle. The geolocation means are configured to locate the vehicle on the geographic map and track the movement or route of the vehicle on the geographic map. For this purpose, the geolocation means may comprise a satellite transmitter / receiver compatible with GPS, Galileo, GLONASS, etc. technologies.

[0038] The geolocation means can also include digital road, topographic, urban maps, etc. For example, SD, HD or OSM cards can be used alone or in combination with each other to extract, in particular, cartographic data. Note that the SD and HD cards provide information on the configuration of the road, the direction of travel of the lanes or even the signage. The SD and HD cards are preferably stored locally in a memory embedded in the vehicle, however it is also possible to store them on a remote server and access them through mobile telephone communication means such as the GSM, 3G, 4G, 5G network. For this purpose, the vehicle can integrate a suitable transmitter / receiver or connect to a mobile terminal such as the driver's smartphone which is equipped with such a transmitter / receiver.

[0039] The OSM map is, for its part, a collaborative map stored on a remote server and accessible by the means of communication previously described. The information contained in the OSM map is entered by the users of the map, it can include precise data on the nature and profile of buildings, signage, pedestrian crossings, etc. It is thus possible to extract cartographic data about the specific features of the environment of the road network in which the vehicle is moving. For example, it is possible to determine whether the vehicle is moving in a speed-regulated zone or near a school, a train station or any other building near which the traffic lanes are shared between motorists, pedestrians, cyclists, etc. For example, in France, there are, in urban areas, 30 zones where the speed is limited to 30 km / h.The OSM map can also list accident zones on a road network.

[0040] In rural areas, it will be possible to determine whether the road crosses an agricultural area in which agricultural vehicles often use the roads on which the vehicle travels or whether the vehicle crosses an accident-prone area.

[0041] According to one embodiment, the capture members 11 may comprise at least one external front camera, the camera is configured to film the road in front of the vehicle. The camera may for example be arranged at the junction between the hood and the front windshield or on the grille of the vehicle. The front camera is configured to determine the configuration of the road, the visibility and also identify obstacles in its recording field but also the traffic signs that are installed on the edges of traffic lanes. In order to recognize the traffic signs that are filmed by the front camera, the security system can integrate a visual recognition algorithm for traffic signs. This type of algorithm is known to those skilled in the art, for example, convolutional neural networks also known by the acronym "CNN" ("convolutional neural networks") include a particular architecture that allows image processing. Among these convolutional neural networks, the neural network known by the acronym "YOLO" ("You Only Look Once") can be used to recognize traffic signs. These visual recognition algorithms are trained with databases of traffic sign images before being used in this application.

[0042] The capture members 11 may also comprise at least one volumetric detection sensor such as a volumetric remote sensing laser known by the acronym “lidar”, or a volumetric detection radar. In one embodiment, the capture members comprise a volumetric remote sensing laser and a volumetric detection radar. Indeed, it is advantageous to correlate the two types of volumetric image to have greater precision. The volumetric detection sensor(s) are generally mounted on the grille of the vehicle. The correlation of the volumetric images and the images captured by the optical camera makes it possible to detect fixed or moving obstacles which are located on the vehicle's travel path or near this path such as on a sidewalk adjacent to this path or even a verge.According to a preferred embodiment, the system also proposes to use a aforementioned visual recognition algorithm to identify the obstacle that is detected. The data from the visual recognition algorithm and the volumetric data can be correlated for example by registration. A modeling algorithm can be used to perform such registration. A Kalman filter type algorithm or equivalent can be used by a person skilled in the art to perform such registration.

[0043] The capture members 11 may also comprise one or more distance sensors which are preferably arranged on the front grille of the vehicle. The distance sensor may be a time-of-flight type sensor and may use ultrasonic technology. It makes it possible to determine the distance with fixed or moving obstacles which are located in front of the vehicle. By correlating the distance data provided by this type of sensor, it is possible to precisely determine the distance between the vehicle and a fixed or moving obstacle which has been detected but also to know whether the obstacle is fixed or moving depending on the movement of the latter in relation to the movement of the vehicle.

[0044] As illustrated in [Fig.l], the system 10 comprises a control unit 12 which is configured to detect a risk situation. The control unit 12 may be an algorithm stored and executed by electronic means on board the vehicle. The control unit 12 may, for example, be integrated at the level of the general computer of the vehicle or be integrated into dedicated electronics. More particularly, a risk situation is detected among a plurality of predetermined accident risk factors 13 which are stored in a memory to which the control unit has access.

[0045] In the embodiment illustrated in [Fig.2], the plurality of risk factors 13 comprises risks of collision 130 with a fixed or moving obstacle. The risks of collision are called "collision" in Figures 2 and 3. The risks of collision can be detected when the distance with a detected obstacle reduces in a determined time. This results in a probability that the trajectories of the obstacle and the vehicle intersect at a given time. According to one embodiment, the reduction in distance in a determined time can be combined with a severity factor making it possible to quantify the seriousness of a possible collision, this severity factor can use vehicle movement data such as speed.Indeed, while the vehicle is moving at a speed of 1 km / h and even if the system detects a very high probability of collision with a third-party vehicle which is in front of the vehicle equipped with the system, such a collision is not likely to cause significant damage, the system may then not trigger an alert. Conversely, when the system detects an average probability of collision with a pedestrian while the vehicle is moving at 70 km / h, an alert is sent because the collision may have significant consequences and take the life of the pedestrian.

[0046] Alternatively, the risk of collision can be detected by triggering automatic emergency braking of the “AEB” type which equips a large number of vehicles currently in circulation. The risks of collision can also be determined using distance sensors located at the front of the vehicle. The control unit 12 can be configured to detect a risk of collision when the distance to an obstacle becomes less than a predetermined threshold value.

[0047] The plurality of risk factors 13 includes overspeed risks 131 which are detected when the vehicle exceeds the speed limits. The overspeed risk is called “overspeed” in Figures 2 and 3. Nowadays, the speed limits of a road are known to most navigators who use a geolocation service. Here, the system 10 can use the onboard navigator of the vehicle, for example, at the central console of the dashboard to extract this information. Alternatively, the system 10 can include wired or wireless connection means with a mobile terminal such as the driver's smartphone to extract data from a navigation application installed on the mobile terminal.

[0048] The connection means are conventional and well known to those skilled in the art, for example of the USB type, for a wired connection and / or of the WIFI / Bluetooth type for a wireless connection.

[0049] Furthermore, the control unit 12 can be configured to compare the vehicle speed data with the map data. The control unit can then detect a so-called “contextual” risk of overspeeding 131 when the vehicle is traveling fast in relation to the environment it is crossing. This situation can be encountered when the vehicle complies with the speed limits but crosses a sensitive area in which there is an establishment that causes significant pedestrian and cyclist traffic, such as a train station, a market or a school. For these purposes, the control unit can correlate the map data of an OSM type map with the vehicle speed which is available in the BUS data in particular which are well known to those skilled in the art.

[0050] The plurality of risk factors 13 further comprises hidden zones 132 in the environment that the vehicle is passing through. The hidden zones 132 are referred to as “hidden zones” in FIGS. 2 and 3. The hidden zones 132 correspond to zones that are not visible to the driver because of an obstacle that may be moving, such as a parked truck or a bus pulled over at a bus stop. In addition, a building, an infra-road or advertising structure that are stationary elements may, depending on the configuration of the road, constitute visual obstacles that may create zones hidden from the driver's vision.

[0051] The control unit 12 is thus configured to determine the hidden zones 132 by comparing the cartographic data using, for example, an OSM type map with respect to the geolocation of the vehicle and the images recorded by the camera.

[0052] In the example illustrated in [Fig.2], the plurality of risk factors 13 comprises risks of changing lanes 133. The change of lane is called in English "lane change". The control unit 12 is configured to detect a change of lane by comparing the map data with the geolocation of the vehicle and the images recorded by the front camera. The map data provided by an HD digital map is sufficiently precise to define whether the driver crosses a signal line delimiting the lane on which he is traveling.

[0053] In the example illustrated in [Fig. 2], the plurality of risk factors 13 includes risks of frontal impact 134 called “inter-distance” in Figures 2 and 3. The risks of frontal impact 134 are in fact particular cases of collisions which can be associated with a line crossing when it is the vehicle which carries the system 10 which veers onto the traffic lane which has an opposite direction by in relation to the direction of travel of said vehicle. Conversely, if it is a third-party vehicle which finds itself in the opposite direction on the traffic lane of the vehicle equipped with the system 10, the situation does not present detection of line crossing. As a result, the risks of frontal impact 134 can be detected in the same way as the risk of collision by combining it in certain cases with the risks of line crossing.

[0054] As illustrated in [Fig.l], the system 10 comprises driver warning means 14 which are configured to warn the driver of the imminence of a risky situation. The triggering of the warning means 14 is determined by the control unit 12 as a function of the imminence of the risky situation.

[0055] According to the system 10 and the method 50 according to the invention, two types of risks can be distinguished: an immediate risk and a foreseeable risk. Generally speaking, the immediate risk is linked to another road user (vehicle, pedestrian, debris on the road) and can correspond to the risk factors of collision 130, frontal impact 134. In the event of detection of an immediate risk, the control unit is configured to actuate the warning means 14 within a second.

[0056] Conversely, a foreseeable risk is more frequently due to the configuration of the road and the environment adjacent to the road. The control unit 12 is configured to actuate the warning means 14 a few seconds before the approach of the risky situation, in particular around ten seconds before the risky situation. This allows the system and method according to the invention to prevent risky situations when approaching an accident-prone intersection in particular.

[0057] In embodiments, the warning means 14 may comprise: - vehicle control equipment equipped with a haptic function such as the steering wheel, the accelerator pedal, the brake pedal, - Visual stimuli such as a light strip, - drawings and icons stored in a memory and displayed on a screen integrated into a passenger compartment of the vehicle such as the screen of the central console of the dashboard of the vehicle, and - sounds emitted by speakers also integrated into the vehicle's passenger compartment.

[0058] The haptic function sends information to the driver that relates to the sensation of touch, generally these are vibrations. But it can also be increased resistance of the accelerator pedal etc.

[0059] The plurality of risk factors thus includes violations of the traffic rules in force in a given territory, such as speeding, crossing a continuous line, but also failure to comply with signs (traffic lights, stop signs, giving way, etc.). For this purpose, the control unit 12 can integrate a database traffic code data of the State in which the vehicle is operating. The control unit 12 is configured to detect a traffic violation based on the vehicle's movement in its environment, which is captured by the capture devices. When a traffic violation is detected, the control unit is configured to activate the alert means 14. For this purpose, the control unit 12 monitors certain data likely to enable the identification of a traffic violation, including speed data measured on the vehicle in relation to map data, signaling data, camera images, etc.

[0060] As illustrated in [Fig.l], the system 10 comprises a module 15 for analyzing the recorded data in order to determine the origin of the risk situation. For this purpose, the analysis module 15 is configured to determine at least contextual parameters 16 of the risk by analyzing the data recorded by the capture devices 11. The contextual parameters 16 are defined arbitrarily and make it possible to contextualize the triggering of the warning in relation to the situation in which the vehicle is located by using parameters linked to the environment, the vehicle and the behavior of the driver. The analysis module 15 can be an algorithm which has access to predetermined contextual parameters 16 stored in one of the memories of the system 10.

[0061] In the embodiment illustrated in [Fig. 2], the contextual parameters 16 determined by the analysis module 15 comprise a parameter relating to a configuration 160 of the road crossed using the road profile data. This parameter 160 is called “road structure” in Figures 2 and 3. As described previously, the recorded data comprise road profile data which are extracted from the recorded images of the road and map data according to the geolocation of the vehicle. The analysis module 15 is configured to extract and correlate this data in order to determine the configuration of the road when the risk factor(s) have been detected.

[0062] As illustrated in Figures 2 and 3, the contextual parameters 16 comprise a parameter defining the behavior of the vehicle when approaching said road configuration, where at least one risk factor has been detected by the control unit 12. The behavior of the vehicle is called “Maneuver” in Figures 2 and 3. It corresponds more precisely to the maneuver that the vehicle was making at the time of detection of the risky situation: was the vehicle going straight ahead? Was it turning, and if so in which direction? The behavior of the vehicle at a time t can be determined using the vehicle movement data such as the vehicle speed, braking, steering wheel angle, turn signals. This maneuver data can be recovered by the analysis module 15 on the bus data of the vehicle and correlated with each other to determine the behavior of the vehicle when approaching said road configuration.

[0063] As illustrated in Figures 2 and 3, the contextual parameters 16 include a regulatory context parameter 162 linked to the configuration of the road crossed by the vehicle. The regulatory context is called “Signaling and areas of interest”. The regulatory context 162 can be determined by the analysis module 15 using the signage data which can come from a visual recognition of the signs which are placed at the edges of the road but also from the cartographic data according to the geolocation of the vehicle as described previously. The analysis module 15 is configured to extract the data from the images and the cartographic data in order to correlate them to determine the regulatory context linked to the configuration of the road crossed.

[0064] The area of ​​interest corresponds to the environment that the vehicle is crossing; this is another contextual parameter 16 that can be associated with the regulatory context as illustrated in Figures 2 and 3. The data on the specificities of the road network environment make it possible to determine whether the vehicle is crossing an area in which certain risk factors and / or certain contextual parameters are more or less likely. As described previously, the specificities of the environment make it possible to know whether the vehicle is moving in a speed-regulated area or near a public place or an establishment around which the traffic lanes are shared to a greater extent by pedestrians, cyclists, etc.The analysis module 15 is configured to extract the map data based on the geolocation of the vehicle in order to determine the environment in which the vehicle is traveling upon approaching the detected risk situation by querying the available map data as described above.

[0065] Among the cartographic data, we can find information on speed-regulated zones and pedestrian crossings. This data makes it possible to define risk factors or contextual parameters that are more or less probable depending on the zone that the vehicle is crossing. This data, correlated with the signaling data 162, makes it possible to limit, in particular, the probability of error in recognizing the risky situation.

[0066] Environmental specificity data may also concern an accident zone in which at least one accident has been recorded over the last ten years; this type of data is accessible in accident databases. Indeed, the public authorities of certain States such as France and the United Kingdom establish this type of map and make it public. As far as France is concerned, they can be found on the official road safety website: https: / / www.onisr.securite-routiere.gouv.fr / en / crash-map.

[0067] As illustrated in Figures 2 and 3, the contextual parameters 16 include a parameter for the presence of obstacles 163 likely to interact with the vehicle. In Figures 2 and 3, this parameter is noted “Type of exo”. The analysis module 15 is thus configured to analyze data from fixed or mobile obstacles that are located on the road network near the vehicle. For this purpose, the analysis module 15 can extract the obstacle data from the recorded images, volumetric images of the road in front of the vehicle recorded by at least one electromagnetic sensor oriented in the direction of travel of the vehicle, but also from distance sensors. The correlation of this data makes it possible to recognize the type of obstacle, for example, pedestrian, cyclist, car, truck or bus but also the distance at which the obstacle is located, whether it is moving or not.Obstacle recognition can be achieved by visual recognition using a suitable algorithm as described in this document.

[0068] The analysis module 15 can be configured to process the data from the distance sensor in order to define the distance but also the movement of the obstacle if this distance changes over time and in relation to the movement of the vehicle according to the recorded data. The images recorded by the camera or the volumetric images can also be used to determine whether the obstacle is moving in relation to the movement of the vehicle which is known by the speed, the geolocation and the appropriate cartographic data. These data can be correlated to determine the presence of an obstacle, its nature and its movement. The obstacle parameters 163 make it possible to define whether the alert is due to an interaction with fixed or moving obstacles such as other road users.

[0069] In embodiments illustrated in Figures 2 and 3, the contextual parameters 16 may comprise position 164 and traffic direction 165 parameters which are respectively noted “Ego Lane” and “Lane Direction” in these figures. The analysis module 15 is configured to determine the position 164 and traffic direction 165 parameters by extracting and correlating the geolocation data, the map data and the images recorded by the camera.

[0070] The cartographic geolocation data makes it possible to define whether the vehicle has crossed a marking line, the camera images can also detect a change of lane. Finally, this same data coupled with the ground signaling data can make it possible to detect whether the vehicle was traveling in the correct direction of travel when one or more risk factors were detected.

[0071] In another embodiment illustrated in [Fig. 3], the contextual parameters 16 may include time parameters 166 such as the time and day at which the risk situation was detected. These time parameters 166 are noted “Time of day” in [Fig. 3]. The analysis module 15 is configured to recover this data, for example, on the vehicle's internal clock, remotely via a telecommunications network such as a telephone or satellite. The time data can be correlated with environmental specificity data. It is thus possible to define increased pedestrian traffic if one passes near a station or a school at times when the surroundings of these establishments are actively frequented.

[0072] In the embodiment of [Fig. 3], the contextual parameters 16 may include meteorological parameters 167 which are noted “Weather” in this figure. The analysis module 15 is configured to extract the meteorological data according to the geolocation of the vehicle from a meteorological data website. Such extraction is known to those skilled in the art. The meteorological data may constitute an important contextual parameter in the event of rainy, snowy, windy conditions, etc. Indeed, depending on the weather conditions, the braking distances or the grip on the road may be considerably modified. Taking this into consideration, adequate behavior on dry roads may become dangerous on wet roads.

[0073] According to the invention, the recording of said data relating to the movement of the vehicle and its environment is carried out using a buffer memory 17 of a determined duration which precedes the detection of the risky situation. The buffer memory 17 may in particular be a few seconds and preferably ten seconds. The analysis module 15 is configured to analyze the data recorded in the buffer memory 17 in order to determine the contextual parameters 16.

[0074] Alternatively, the analysis module 15 may be a trained model of the “video-to-text” type. This type of trained model is part of the family of LLM (Large Language Model) models. Typically, this model takes as input the video recorded by the front camera of the vehicle (frame sequence), and extracts from it, by visual recognition, the important information which here corresponds to the contextual parameters 16. Following the extraction of the contextual parameters, the model describes the risky situation which has just been detected. We can cite for example, in this category of trained models the “BLIP-like models”, and in particular the “VideoBLIP” version which is an augmented version of “BLIP-2”, able to process videos (and no longer only images like “BLIP-2”).

[0075] Generally speaking, when the system 10 or the method 50 according to the invention correlates several data to determine a risk factor or a contextual parameter, the information from several sources (mapping data or camera images) is compared with each other in order to increase the probability that the risk factor or the contextual parameter has been correctly detected. In the event of a contradiction between data from two different sources, the most probable information is retained, in particular, by using a third criterion. For example, in case of conflict of overspeed analysis between the signs detected by the camera and map data from the vehicle's browser, if the data on specific features of the area, in particular from the OSM map, indicate the presence of a work zone with a specific speed limit on the section of road in question, the data from the traffic sign captured on video will then be favored over the data from the browser. Alternatively, according to another embodiment, in the event of a contradiction between data from two different sources, if the probabilities between the two sources are equal or below a predetermined threshold, this data may not be used.

[0076] As illustrated in [Fig.l], the system 10 comprises an advisory module 18 configured to produce driving advice when the origin of the risky situation is related to the driver's behavior. In this situation, depending on the risky situation that is detected and the contextual parameters that have been determined, the advisory module 18 is configured to produce advice for the driver. The objective of the advice is to improve the driver's behavior when approaching the risky situation that has been detected or a similar risky situation that is in another location.

[0077] This is why the system 10 is configured to give driving advice relating to the situation that it has experienced in a dedicated application 19 which can be embedded in the main console of the dashboard or in a mobile terminal such as the driver's smartphone.

[0078] In practice, as soon as the warning means 14 are triggered, the driver has the possibility of accessing advice relating to the risky situation which has been detected. Advantageously, the advice is formulated in such a way as to help the driver to better understand the situations which he has encountered (“remember to brake before a pedestrian crossing”, “reduce your speed in such a complex area because…” etc.).

[0079] It is not excluded that several pieces of advice may be given for a given risk situation.

[0080] [Fig.4] illustrates an embodiment of the structure 20 of a board issued by the system 10 and the method 50 according to the invention. The structure 20 of the board may comprise a first part 200 which introduces the risk situation which has been detected. In practice, the introduction is determined according to the contextual parameters 16 which have been determined by the analysis module 15. In the example of [Fig.4], the first part 200 bears the mention “Source of risk” which is defined by the contextual parameters 16 which have been determined. In this example, the risk situation which has been detected is a blind bend.

[0081] As illustrated in [Fig.4], the structure 20 of the advice may comprise a second part 201 which provides advice on the behavior to be adopted in the risky situation. which has been detected. The behavior to be adopted is noted "Expected behavior" in [Fig.4] and in this example, the driver is advised to slow down when approaching a blind bend. The structure 20 of the advice may include a third part 202 which has an educational purpose since it indicates the reasons why this behavior is preferable when approaching such a situation. The third part 202 is marked "Reason for adopting this advice".

[0082] In order to formulate advice, the advice module 18 is configured to receive the risk situation which has been detected by the analysis module 15 according to the determined contextual parameters 16. With this information, the advice module 18 can formulate the first part 200 of the structure 20 of the advice.

[0083] In order to formulate a complete advice, the system 10 may comprise an advice database 21 which comprises advice on behavior to adopt which is classified according to the determined contextual parameters 16. Depending on the detected risk situation, each advice 201 has a probability P of relevance, and the advice having the greatest probability P of relevance is selected by the advice module 18 with a view to being integrated into the structure of the advice. Each predetermined advice is associated with N parameters corresponding to the accident risk factors and to the contextual parameters 16, the advice with which the greatest number of parameters detected in a risk situation are associated corresponds to the advice which has the greatest probability P of being proposed to the driver.

[0084] The advice database 21 may also include reasons why the advice on the behavior to be adopted is suggested to the driver. These reasons are also classified according to a probability P in relation to the advice to be adopted and the detected risk situation. The reasons having the greatest probability P of relevance are selected by the advice module 18 in order to formulate advice using the structure 20 illustrated in [Fig.4]. The probability P of each advice and / or reason may be determined by associating each of them with risk situations but also with contextual parameters 16. The advice grouping the most contextual parameters 16 determined may thus be considered as the most relevant.

[0085] The advice database 21 may also include driving rules such as the highway code on French territory. These rules make it possible to define whether among the determined contextual parameters 16 some correspond to a violation of a traffic rule in force on the territory in which the vehicle is traveling.

[0086] The advice module 18 may be an algorithm or an application configured to compare the input data (detected risk situation) and the data contained in the database 21 in order to formulate advice whose structure 20 is illustrated in [Fig.4],

[0087] The database 21 may be a coded database in which, unless an operator intervenes, the data are listed in a fixed manner. This type of database makes it possible to make a comparison of data to data according to various data weighting criteria, here for example, the number of contextual parameters 16.

[0088] Alternatively, the database 21 may be a trained model of the “Question-Answering” type. These models can answer a question from a given context, in this case in the case of the invention, the question would present the risky situation which has been previously defined using the contextual parameters 16 and would consist of asking the model to give advice to avoid taking risks such as: “In this risky situation, what advice should be given to avoid taking risks?”.

[0089] Trained models for generating text are particularly relevant for generating such advice, including "GPT". For the advice generated by such a model to be effective, it is useful to limit the number of lines that the response generated by the model must contain.

[0090] The advice module 18 may also comprise an algorithm generating explanatory diagrams of the risky situation that has been detected. For this purpose, it is possible to use an image generative algorithm such as “StyleDrop” which are capable of generating from a text, in this case the advice already formulated by the advice module 18, images according to a predefined style for example a cartoon style, sketch etc.

[0091] The advice is thus formulated by the advice module 18 in relation, on the one hand, to the risky situation detected and defined by the analysis module 15 in particular as a function of the contextual parameters 16, and on the other hand, with the database 21. As described previously, the advice is sent to the driver via the application 19. The application 19 is configured to display the advice and to ask the driver whether the advice is relevant to the risky situation he has encountered. The driver can then respond on the human-machine interface on which the application 19 is executed.

[0092] The advice module 18 is configured to modulate the relevance of the advice based on the driver's feedback regarding the relevance of the advice. For example, if the driver judges the advice submitted to him to be relevant in relation to the risky situation he encountered, the relevance of this advice in this situation is reinforced within the database 21. Conversely, when the driver judges that the advice that was formulated is not relevant, the relevance of this advice in the risky situation that was detected will be reduced within the database 21.

[0093] The system 10 is configured to record each advice consulted by the driver. For this purpose, the system 10 comprises a memory 22 to which the advice module 18 has access and which receives feedback from the driver. When the driver again encounters a risky situation for which he has already consulted advice, his driving attitude (speed, information gathering via gaze, reaction time, etc.) is compared, on the one hand, to the information provided by the advice to check whether he follows the advice, and on the other hand, to his driving attitude which was previously recorded, the system 10 and the method 50 are thus configured to monitor the improvement or not of the driver's driving behavior over time and with regard to the advice he consults. In order to evaluate the information gathering of the driver via his gaze, the system 10 may comprise a camera arranged in the passenger compartment and oriented towards the driver.

[0094] According to one embodiment, when the analysis module 15 evaluates an improvement in the driver's behavior for a risky situation for which the latter has viewed advice, the improvement is taken into account when actuating the warning means 14. The improvement in the driver's behavior can be recorded by the analysis module 15 in relation to a personalized profile 23 which is stored in a memory integrated into the system 10. The memory 22 can be used for this purpose or another memory dedicated to storing the driver's history.

[0095] According to one embodiment, when the driver's behavior improves over the course of warnings for the same risky situation that has been detected, the analysis module 15 is configured to modulate, through the personalized profile 23, the actuation of the warning means 14. For example, the number of activations or the intensity of the activations of the warning means 14 can be reduced. The system 10 can also reduce the frequency of activation of the warning means 14 when approaching a risky situation relating to a road configuration that is frequently traveled by the vehicle; the activation of the warning means 14 can be reduced to one pass out of two, for example.

[0096] The system 10 is configured to adapt the activation of the warning means 14, on the one hand, to adapt to the evolution of the driver's behavior, and on the other hand, to avoid becoming invasive when improving the driver's driving behavior.

[0097] The adaptation of the activation of the warning means can be carried out by learning algorithms for the behavior of the driver known as “reinforcement learning”. “Reinforcement learning” is a general method in the field of machine learning which consists of training an agent by defining a score linked to the actions carried out by the agent and by seeking to optimize the values ​​of this score. Applied to the invention, “reinforcement learning” makes it possible to add weightings (positive or negative) to the result to the advice module 18 which gives advice and / or activates an alert. If the advice is well received by the driver, then a positive weighting is added to it to encourage the advice module 18 to act in the same way when the risky situation arises again. This principle can be transposed to the warning means, when the analysis module 15 notes that the risk level decreases each time the driver encounters a risky situation for which he has already received advice, the alerts can then be reduced in frequency and / or intensity. In particular, if the risk level continues to decrease or at least no longer increases significantly, the analysis module 15 tends not to give another alert for such behavior associated with such a risky situation.

[0098] Conversely, if the driver systematically ignores the warnings sent by the advice module 18 on the application 19, for example, for the same road configuration identified and located at a specific place, the advice module 18 no longer offers advice and reduces the number of activations of the warning means 14, this so that the advice and warnings do not become too frequent for the driver and so that he is not pushed to deactivate the system 10 in his vehicle.

[0099] The advice can still be stored in the memory 22 in the form of coaching for possible viewing by the driver, but will no longer be systematically offered to the driver when the driver encounters this risky situation.

[0100] This will reduce the number of warnings sent to the application 19 for risky situations for which the driver is making progress in his behavior as he approaches them. The system 10 is thus configured to focus the warnings and advice more on particularly dangerous risky situations. The system 10 is thus configured to overcome the habituation effect of the warnings.

[0101] As illustrated in Figures 1 and 5, the invention also relates to the method 50 for evaluating situations at risk of accidents for a vehicle traveling on a road network in a given environment, the method 50 comprising a step 51 for monitoring the progress of the vehicle on the road network and in the environment in which the road network is integrated and the vehicle is moving. As described previously, the monitoring step 51 is carried out by recording different types of data such as vehicle movement data, in particular the speed of the vehicle and its geolocation, road profile data that the vehicle is crossing, road signage data for the roads that the vehicle is crossing, specific features of the environment of the road network in which the vehicle is moving, fixed or mobile obstacles that are located on the road network near the vehicle.

[0102] As illustrated in [Fig.l], the system 10 performs the tracking step 51 using the capture members 11 and the control unit 12 and the buffer memory 17 in which the data is stored.

[0103] According to one embodiment, the buffer memory 17 stores said data for a determined duration, called the buffer duration, which may be a few seconds, and preferably the buffer duration is ten seconds. Therefore, when a risky situation is detected, the method 50 has the ten seconds of recording which precede the detection of the risky situation to analyze the contextual parameters 16 which are linked to the risky situation.

[0104] As illustrated in [Fig.5], the evaluation method 50 comprises a step 51 of detecting a risk situation. The risk situation is detected by monitoring, through the recorded data, a plurality of accident risk factors 13. As previously described, the plurality of risk factors 13 may include traffic violations, risks of collision 130 with a fixed or moving obstacle, risks of overspeeding 131, risks of hidden zones 132 in the environment that the vehicle is crossing, risks of changing lanes 133, and risks of frontal collision 134 on the same traffic lane. The manner in which the risk factors are recognized has been described previously.

[0105] The method 50 detects a risky situation by comparing, in real time, the recorded data with the plurality of risk factors. According to the system 10, it is the control unit 12 which implements this comparison work. For example, in the event of a detected line crossing, if this line is recognized by the images recorded by the visual recognition algorithm, this situation is compared to the highway code and if it is contrary to the latter, a traffic violation is detected. In this case, two risk factors would be detected by the method, on the one hand, a line crossing 133 and on the other hand a traffic violation.

[0106] As illustrated in Figures 1 and 5, when one or more risk factors are detected simultaneously, a risky situation is detected, the method 50 comprises a step 53 of warning the driver. According to one embodiment, the warning 53 may comprise visual stimuli emitted in particular by a light strip, vehicle control equipment equipped with a haptic function such as the steering wheel, the accelerator pedal, the brake pedal, drawings and icons stored in a memory and displayed on a screen integrated into a passenger compartment of the vehicle, and sounds emitted by loudspeakers also integrated into the passenger compartment of the vehicle. Of course, according to the invention, a combination of several visual, sound, and haptic stimuli may be envisaged to warn the driver of the imminence of a risky situation.

[0107] When a risky situation is detected, the method 50 comprises a step 54 of analyzing the recorded data in order to determine the origin of the situation. risk. In particular, it is the data recorded in the buffer 17 which are analyzed. The analysis 54 is carried out at least by determining the following contextual parameters 16 of the risk: - a configuration of route 160 crossed using road profile data, - a behavior 161 of the vehicle when approaching said road configuration using the vehicle movement data, - a regulatory context 162 linked to the configuration of the road crossed, the context being determined by signaling data - the environment 162 that the vehicle is crossing, the data on the specificities of the road network environment make it possible to determine whether the vehicle is crossing an area in which certain risk factors are more or less probable, and - the presence of obstacles likely to interact with the vehicle.

[0108] The analysis of the obstacle data makes it possible to define whether the warning is due to an interaction with fixed or moving obstacles such as other road users, and if so, whether the risky situation is at least partly due to the driving attitude of the driver or to other road users. As described above, the analysis module 15 is configured to analyze the recorded data in order to determine the contextual parameters 16. These contextual parameters 16 allow the method 50 to define the circumstances of the detection of the risky situation and thus to adapt the advice it formulates according to these circumstances.

[0109] In an embodiment illustrated in [Fig.2], the contextual parameters 16 may comprise position parameters 164 and direction of travel 165 of the vehicle on lanes of the road. As described previously, the parameters 164, 165 are determined by the analysis module 15 as described previously using certain data recorded in the buffer memory 17.

[0110] In an embodiment illustrated in [Fig.3], the contextual parameters 16 may comprise temporal parameters 166. The temporal parameters 166 are determined as described previously and may be correlated with the specificity parameters of the environment such as the nature of the buildings near the area crossed by the vehicle. The method 50 can then identify a significant probability of pedestrian traffic in said area when the vehicle passes, for example, near a school at the time when the students leave.

[0111] In one embodiment shown in [Fig.3], the contextual parameters 16 may include weather parameters 167. The weather data is recorded during the tracking step 51 as previously described. The weather parameters 167 make it possible to adjust the advice in the event of adverse weather. favorable (rain, hail, snow, fog, etc.). Indeed, the driving attitude can be radically different in the event of heavy rain or thick fog. Such a nuance makes it possible to refine the formulation of driving advice.

[0112] As illustrated in [Fig.5], the method 50 comprises a step 55 of producing driving advice. In particular, the driving advice identifies the detected risky situation and recommends driving behavior in order to reduce the risks incurred by the vehicle when approaching said situation. The driving advice is formulated in particular when the origin of the risky situation relates to the behavior of the driver, the advice being produced according to the determined contextual parameters.

[0113] As described previously, the production step 55 is carried out by the advice module 18. According to an embodiment illustrated in [Fig.l], the advice is produced from the risk situation defined by contextual parameters 16 which have been determined and from an advice database 21. The advice database 21 may include predefined recommended driving behaviors when approaching the determined risk situation as well as reasons why said behaviors are recommended in the detected risk situation.

[0114] According to one embodiment, the advice can be produced by comparing said determined contextual parameters with recommended behaviors and reasons stored in the database 21. As described previously, each recommended behavior and reason for the advice are associated with a probability P of relevance in relation to a detected risk situation. The recommended behavior and the reasons for the advice which have the greatest probability P are selected in order to produce the advice. The probability P of relevance can be defined according to the contextual parameters 16 and the risk situation which are detected.

[0115] According to an embodiment illustrated in Figures 1 and 5, the method 50 comprises a step 56 of sending the advice to the driver. As described previously, the advice is sent to the driver via the application 19. The latter can be stored and executed on a terminal such as a smartphone or the central console of the vehicle. The terminal must have a human-machine interface to allow the user to provide feedback on the relevance of the advice. Thus, the method 50 can comprise a feedback step 57 on the relevance or otherwise of the advice given to him depending on the risky situation he encountered. This feedback step 57 is taken into account by the method 50 which can comprise, according to one embodiment, a step 58 of adjusting the relevance P of the driving advice depending on the driver's feedback as described previously.As described above, it is the advice module 18 which carries out the adjustment step 58.

[0116] In parallel, according to an embodiment illustrated in figures 1 and 5, when a advice is consulted by the driver, the method 50 may comprise a step 59 of recording the advice in connection with the situation encountered by the driver. This recording 59 is carried out as described previously, in particular, the recording step 59 is carried out by the analysis module 15 and makes it possible to monitor the behavioral evolution of the driver when approaching risky situations which are regularly detected on the journeys made by the driver.

[0117] Depending on the evolution of the driver's behavior when approaching a risky situation that has already been detected in the past, the method 50 may comprise a step 60 of adapting the emissions of the warnings 53 as well as the production 55 of the advice. In [Fig.5], the influence of the adjustment 60 on the warnings is symbolized by a dotted arrow which connects the bubbles representing the two steps. For example, in the event of an improvement in the driver's behavior when approaching a risky situation to be determined, if the method 50 and the system 10 note, in particular through the contextual parameters 16, a reduction in risk-taking, the emission of the warnings 53 may be reduced in frequency or intensity. The method 50 is thus configured to modulate the warnings 53 according to the driver's evolution and the dangerousness of the situations that he is likely to encounter.By reducing the intensity or frequency of warnings for situations that present a moderate risk, process 50 becomes more effective when addressing situations where the risk of accidents is greater.

[0118] According to another embodiment, the method may also comprise a step 61 of adapting the advice that is sent to the driver. This step 61 also uses the monitoring of the behavioral evolution of the driver to adapt the frequency at which advice is sent to the driver for a determined risky situation that may be encountered frequently by the driver. In [Fig.5], the influence of the adaptation 61 of the advice is symbolized by a dotted arrow that connects the bubbles representing the sending of the advice 56 and the adaptation 61 of the latter according to the behavior of the driver.The objective of this modulation is the same as the adaptation step 60 of the warnings, that is to say, it cannot over-solicit the driver when approaching situations for which his behavior is good or is improving, but inform him more effectively for risky situations that he encounters infrequently or which present a very high risk of accident.

[0119] Although the advice is not sent to the driver for situations for which his behavior is appropriate, the method 50 can store this advice and make it available to the driver, in particular, in the application 19.

[0120] The method 50 is also configured to inhibit advice that is systematically ignored by the driver. Inhibiting advice, in this case, is a special case of adaptation 61 of the sending of advice. The inhibition of advice may not be definitive, it may be proposed to the driver less frequently.

Claims

1. Claims Method for evaluating (50) situations involving an accident risk for a vehicle in circulation on a road network in a given environment, the method comprising: - monitoring (51) the evolution of the vehicle on the road network and in the environment in which the road network is integrated and the vehicle moves, the monitoring (51) is carried out by recording: vehicle movement data, in particular the speed of the vehicle and its geolocation, road profile data that the vehicle crosses, road signage data that the vehicle crosses, data on the specificity of the environment of the road network in which the vehicle moves, fixed or mobile obstacles that are located on the road network near the vehicle; - the detection (52) of a risky situation among a plurality of accident risk factors (13) comprising: traffic violations, risks of collision (130) with a fixed or moving obstacle, risks of overspeeding (131), risks of hidden zones (132) in the environment that the vehicle is crossing, risks of changing lanes (133), and risks of frontal collision (134) on the same traffic lane, a risky situation being detected by comparing, in real time, the recorded data with the plurality of risk factors (13); - warning (53) of the driver when a risky situation is recognized; - the analysis (54) of the recorded data which precede the detection (52) of the risk situation in order to determine an origin of the risk situation, the analysis (54) being carried out at least by determining contextual parameters (16) of the risk: • a configuration of the road (160) crossed using the road profile data, • a behavior (161) of the vehicle when approaching said road configuration using the vehicle movement data, • a regulatory context (162) linked to the configuration of the road crossed, the context being determined by signaling data • the environment that the vehicle is crossing, the data on the specificities of the road network environment make it possible to determine whether the vehicle is crossing an area in which certain risk factors are more or less probable, • the presence of obstacles (163) likely to interact with the vehicle, the analysis of the obstacle data making it possible to define whether the warning is due to an interaction with fixed or mobile obstacles such as other road users; and the production (55) of driving advice based on the determined contextual parameters (16), the driving advice identifies the detected risky situation and recommends driving behavior in order to reduce the risks incurred by the vehicle when approaching said situation.

2. Evaluation method (50) according to claim 1, in which the production of the advice (55) is carried out from the risk situation defined by contextual parameters (16) which have been determined and from an advice database (21) which comprises: - predefined recommended driving behaviors when approaching the determined risk situation, and - reasons why said behaviors are recommended.

3. Evaluation method (50) according to claim 2, wherein, the production (55) of the advice is carried out by comparing said determined contextual parameters (16), with recommended behaviors and reasons stored in the database, the recommended behaviors and said reasons which correspond to said contextual parameters (16), are associated with the detected risk situation (16) in order to produce advice.

4. Evaluation method (50) according to one of claims 1 to 3, which comprises adjusting (58) the relevance of the driving advice in depending on the driver's reaction when the warning (53) is triggered for a specific risk situation or when the driver provides feedback on the relevance of the driving advice after the method (50) has questioned the driver.

5. Evaluation method (50) according to one of claims 1 to 4, which comprises adapting the emissions of the warnings (53) and the production (55) of the advice depending on the behavior of the driver when approaching a risky situation which has already been detected in the past.

6. Evaluation method (50) according to one of claims 1 to 5, in which the recording of said data relating to the movement of the vehicle and its environment is carried out using a buffer memory of a determined buffer duration which precedes the detection of the risky situation, the buffer duration can in particular be a few seconds and preferably ten seconds, thus when a risky situation is detected, the method (50) analyses the data recorded in the buffer memory.

7. Evaluation method (50) according to one of claims 1 to 6, wherein said data comprises road profile data which are extracted from the recorded images of the road and the map data according to the geolocation of the vehicle, the recorded images, the geolocation and the map data being correlated to determine the contextual parameter of configuration of the road (160).

8. Evaluation method (50) according to one of claims 1 to 7, wherein said data comprises vehicle maneuver data including speed, braking, steering angle, and indicator data, the maneuver data being correlated to determine the contextual behavior parameter (161) of the vehicle upon approaching said road configuration.

9. Evaluation method (50) according to one of claims 1 to 8, wherein said data comprises digital signaling data which are extracted from the map data according to the geolocation of the vehicle and visual signaling data which are extracted from images recorded by the front camera of the vehicle, the data extracted from the images and the map data are correlated to determine the regulatory context (162) linked to the configuration of the road crossed.

10. Evaluation method (50) according to one of claims 1 to 9, in which said data comprises specificity data of the environment that the vehicle is crossing, the specificity data is extracted from cartographic data according to the geolocation of the vehicle, the specificity data comprises information on the types of buildings near roads, regulated speed zones, pedestrian crossings, the specificity data makes it possible to determine particularities of the environment that the vehicle is crossing which are linked to the regulatory contextual parameter (162).

11. Evaluation method (50) according to claim 10, in which the contextual parameters (16) comprise temporal parameters (166) which are determined by temporal data recorded during the monitoring step (51), the temporal data providing information on the time and day at which the risk situation was detected, the temporal data are, in particular, correlated with the specificity data of the environment.

12. Evaluation method (50) according to one of claims 1 to 11, wherein, said data comprises data of fixed or mobile obstacles which are located on the road network in the vicinity of the vehicle, the obstacle data are extracted from the recorded images, volumetric images of the road in front of the vehicle recorded by at least one volumetric detection sensor oriented in the direction of travel of the vehicle and distance data recorded by a distance sensor, the data extracted from the images, volumetric images and distance are correlated to determine the obstacle parameter (163).

13. Evaluation method (50) according to one of claims 1 to 12, in which the contextual parameters (16) comprise position parameters (164) and direction of travel (165) of the vehicle on lanes of the road, these parameters are determined by traffic data comprising a position of the vehicle on lanes of the road, and a direction of travel of the vehicle on this lane, the traffic data are extracted from the geolocation, the cartographic data and the images recorded by the camera, these traffic data make it possible to define whether the vehicle has crossed a marking line on the ground and / or whether the vehicle was traveling in the correct direction of travel.

14. Evaluation method (50) according to one of claims 1 to 13, in which the contextual parameters (16) comprise meteorological parameters (167) which are determined by meteorological data recorded during the tracking step (51), the meteorological data are extracted according to the geolocation of the vehicle from a meteorological data website.

15. Evaluation method (50) according to one of claims 1 to 14, wherein the plurality of accident risk factors (13) comprises - risks of collision (130) with a fixed or mobile obstacle which are detected when the distance with an obstacle is reduced in a determined time, - risks of overspeed (131) which are detected when the vehicle exceeds the speed limits or when the vehicle is traveling fast in relation to the environment in which the vehicle is moving, - risks of hidden zones (132) in the environment crossed by the vehicle, the hidden zones are determined by comparing the map data with the geolocation of the vehicle and the recorded images, - risks of changing lane (133), the change of lane is determined by comparing the map data with the geolocation of the vehicle and the recorded images,and - risks of frontal collision (134) on the same traffic lane which are determined by the recorded images and distance sensors located in front of the vehicle and map data.,

16. Evaluation method (50) according to one of claims 1 to 15, in which the warning (53) comprises visual stimuli emitted in particular by a light strip, vehicle control equipment equipped with a haptic function such as the steering wheel, the accelerator pedal, the brake pedal, drawings and icons stored in a memory and displayed on a screen integrated in a passenger compartment of the vehicle, and sounds emitted by loudspeakers also integrated in the passenger compartment of the vehicle.

17. Land vehicle, in particular an automobile, characterized in that it implements the method (50) for evaluating situations at risk of accidents in a vehicle circulating on a road network in a specific environment which is defined according to claims 1 to 16.

18. System for evaluating (10) situations involving an accident risk for a vehicle in circulation on a road network in a given environment, the system comprising: - capture members (11) of the vehicle's movement on the road network and in the environment in which the router network is integrated, the capture members (11) being configured to record vehicle movement data, in particular the speed of the vehicle and its geolocation, road profile data that the vehicle is crossing, signaling data of the roads that the vehicle is crossing, data on the specificity of the environment of the road network in which the vehicle is moving, fixed or mobile obstacles that are located on the road network near the vehicle; - A control unit (12) which is configured to detect a risk situation among a plurality of accident risk factors (13) comprising: traffic violations, risks of collision (130) with a fixed or moving obstacle, risks of overspeeding (131), risks of hidden zones (132) in the environment that the vehicle is crossing, risks of changing lanes (133), and risks of frontal collision (134) on the same traffic lane, the control unit (12) being configured to detect a risk situation by comparing, in real time, the recorded data with the plurality of risk factors (13); - driver warning means (14) which are configured to warn the driver of the imminence of a risky situation, - an analysis module (15) of the recorded data which precede the detection of the risky situation in order to determine the origin of the risky situation, the analysis module (15) being configured to determine at least parameters contextual (16) of the risk: • a configuration of the road (160) crossed using the road profile data, • a behavior of the vehicle (161) when approaching said road configuration using the vehicle movement data, • a regulatory context (162) linked to the configuration of the road crossed, the context being determined by signaling data • the environment that the vehicle is crossing, the specific data of the road network environment makes it possible to determine whether the vehicle is crossing an area in which certain risk factors are more or less probable, • the presence of obstacles (163) likely to interact with the vehicles, the analysis of the obstacle data makes it possible to define whether the warning is due to an interaction with fixed or mobile obstacles such as other road users; and an advisory module (18) configured to produce driving advice based on the determined contextual parameters (16), the driving advice identifies the detected risky situation and recommends driving behavior in order to reduce the risks incurred by the vehicle when approaching said situation.

Images