JP7119365B2 - Driving behavior data generator, driving behavior database - Google Patents

Description

The present invention provides a driving behavior data generation device that generates lane-level driving behavior data by associating vehicle driving conditions with environmental conditions and vehicle conditions, and a driving behavior support device that expresses a two-dimensional spatial distribution on a road surface. It relates to a driving behavior database used when executing a process to do.

In recent years, as a driving behavior support device, a system that creates road information using probe data has been proposed.

Japanese Patent Laid-Open No. 2002-200003 describes collecting information from on-vehicle sensors to automatically generate a lane-level road network. In Patent Document 1, unlike the conventional navigation map composed of nodes and links, the center line is detected for each lane, and the connection relation for each lane is automatically generated at intersections. The automated driving system uses this information as a clue to design its own vehicle route to the destination.

Patent Literature 2 describes analyzing travel history data and creating an optimum speed for traveling in a specific section connecting intersections with traffic lights. Patent Document 2 aims to reduce the possibility of stopping at a red light from the viewpoint of fuel saving, and creates a speed zone suitable for passing from statistical data without directly measuring the timing of signal switching.

By the way, in the automatic driving/driving support system, it is important to generate human-like behaviors in order to give the passengers peace of mind and comfort.

Non-Patent Document 1 describes the use of the concept of a potential map in order to make a human-like route plan. That is, Non-Patent Document 1 integrates the actual risk potential generated based on the map information and the trajectory of the preceding vehicle and the empirically obtained latent risk potential for the traffic participants to create a potential map around the own vehicle. is generated, and based on that, an optimal route is generated.

JP 2015-4814 A JP 2011-153834 A

"Chunzhao Guo, et al.,“Human-like Behavior Generation for Intelligent Vehicles in Urban Environment based on a Hybrid Potential Map,”Proc.of 2017 IEEE Intelligent Vehicles Symp.(IV),pp.197-203,2017."

However, only the lane information described in Patent Literature 1 can be used to know only the center line along which the vehicle should travel and the range in which the vehicle can travel. For example, when making a right turn at a large intersection, you have to wait for the oncoming traffic to break off within the intersection.

In addition, course selection at intersections is not necessarily the same everywhere, and road surface paint guidance is not always appropriate for that car.

Furthermore, even if the road is a single road, there are road sections that are locally used as multiple lanes in order to efficiently turn right and left.

In other words, the greater the amount of information required for driving, the more it becomes possible to create an automated driving system with driving behavior that is closer to that of a human being, as in Non-Patent Document 1, which can improve the sense of security and comfort of the occupants. However, it does not specifically show how information is collected and how the collected information is used to generate a driving behavior database.

Patent document 2 can be seen as collecting driving behaviors for properly driving on local roads, but the average vehicle speed between intersections is insufficient for an automatic driving system, and and information about the appropriate speed along that route is sought.

Efficient driving on public roads requires not only a lane-level road network, but also local rules associated with it, knowledge suitable for the surrounding environment, and empirical know-how.

SUMMARY OF THE INVENTION An object of the present invention is to obtain a driving action data generation device capable of generating driving action data in consideration of local rules, knowledge suitable for the surrounding environment, and empirical know-how.

In addition to the above purpose, it is also an object to obtain a driving behavior database that can support automatic driving based on driving behavior close to humans by being used in a driving behavior support device that expresses a two-dimensional spatial distribution on a road. is.

The driving action data generation device of the present invention includes an acquisition means for acquiring driving information related to driving of the vehicle from an in-vehicle sensor installed in the vehicle, and the driving information acquired by the acquisition means is stored on a known driving route. Determination means for determining a travel history including a vehicle position and a travel trajectory for each lane of a travel road composed of a plurality of lanes by collating with structural information; A classifying means for classifying each lane of a traveling road using at least one of vehicle conditions related to motion performance and surrounding environmental conditions when the vehicle is traveling as a classification requirement; generating means for generating, for each classification requirement classified by the classification means, driving action data used when executing guidance relating to at least one of possible routes and speeds that the vehicle can take, as a basic configuration . .
In the first invention, in addition to the basic configuration, the guidance on the route expresses a road area centered on the lane of the travel route as a two-dimensional grid map, and the two-dimensional grid map is displayed on the lane unit of the travel route. It is characterized by the generation of a spatial distribution that projects the information on the travel history aggregated in .
In addition to the basic configuration, the second invention relates to surrounding environment recognition means for recognizing the surrounding environment including the presence or absence of facilities from the driving information acquired from the vehicle-mounted sensor, and the surrounding environment recognized by the surrounding environment recognition means. Exceptional travel determining means for determining, based on the information and the travel trajectory of the vehicle, whether or not the travel trajectory includes at least one of lane change and obstacle avoidance behavior. A running locus determined as an exceptional run is excluded from classification targets by the classifying means.
Furthermore, in addition to the basic configuration, the third invention extracts irregular driving based on the driving operation including sudden braking and sharp steering based on the driving information and statistical data of the operation status of the safety system installed in the vehicle. The vehicle is characterized by further comprising irregular travel extracting means, wherein the travel locus extracted as the irregular travel is registered as risk information for classification requirements by the classifying device.
In addition to the basic configuration, a fourth invention further comprises moving object information detection means for detecting moving object information from infrastructure sensors including surveillance cameras installed along the travel route, and said moving object information detection means. is added to the information on the traveling history of the vehicle passing through the traveling road in the area where the infrastructure sensor is installed.

According to the present invention, the acquisition means acquires the driving information related to the driving of the vehicle from the in-vehicle sensor mounted on the vehicle.

The determining means compares the driving information acquired by the acquiring means with known structure information of the traveling road, and determines the traveling history including the vehicle position and the traveling trajectory for each lane of the traveling road composed of a plurality of lanes. do. For example, in the case of a road with two or more lanes, the travel history is determined for each lane on which the vehicle is traveling.

The classification means classifies the information about the travel history determined by the determination means in units of lanes of the travel road using at least one of the vehicle condition and the environmental condition as classification criteria. As a result, the travel history information is more detailed than when the travel histories are aggregated regardless of lanes.

For example, a driving support device guides a possible route or speed of a vehicle when the vehicle is traveling on a road in a specific area.

The generation means generates driving action data used when executing the guidance for each classification requirement classified by the classification means. Since the driving behavior data is obtained from the driving history information for each lane of the driving road, it is possible to provide appropriate guidance.

In the first invention , in addition to the basic configuration, the guidance on the route expresses a road area centered on the lane of the travel route as a two-dimensional grid map, and the two-dimensional grid map is displayed on the lane unit of the travel route. It is characterized by the generation of a spatial distribution that projects the information on the travel history aggregated in .

By showing the spatial distribution, it is possible to visually guide the route.

In the present invention, the speed guidance is characterized in that the speed distribution of the vehicle along the road in the specific area is generated .

By showing the spatial distribution, it is possible to visually guide the speed.

In addition to the basic configuration, the second invention includes surrounding environment recognition means for recognizing the surrounding environment including the presence or absence of facilities from the driving information acquired from the vehicle-mounted sensor, and information on the surrounding environment recognized by the surrounding environment recognition means. and an exceptional travel determination means for determining, based on the travel locus of the vehicle, whether or not there was an exceptional travel including at least one of a lane change on the travel route and an obstacle avoidance action with respect to the travel locus, A travel locus determined as an exceptional travel is excluded from classification targets by the classification means.

Exceptional driving that includes at least one of lane change and obstacle avoidance behavior may be inappropriate as driving behavior data, so it is excluded from classification targets by the classification means.

In addition to the basic configuration, the third invention is an irregular driving that extracts irregular driving based on statistical data of driving operations including sudden braking and sudden steering based on the driving information and the operation status of the safety system installed in the vehicle. The vehicle is characterized by further comprising travel extracting means, wherein the travel locus extracted as irregular travel is registered as risk information for classification requirements by the classifying means.

Driving operations, including sudden braking and steering, and the operation of safety systems installed in the vehicle may cause irregular driving, but there are also cases where driving habits and the safety system are constantly activated. Therefore, irregular driving is extracted based on statistical data of driving operation and safety system operation.

The running locus extracted as the extracted irregular running is registered as risk information of classification requirements by the classification means.

The present invention further comprises section setting means for setting road sections for classifying the information related to the travel history determined by the determination means, wherein the section setting means determines points where the number of lanes on the road increases or decreases, It is characterized by the setting of sections of roads that are specialized for turning points including merging and branching points.

The section setting means sets sections of the traveling road by specializing in changing points including points where the number of lanes on the traveling road increases or decreases, and points where the traveling road merges and diverges, so that the changing points of the traveling road can be set more precisely. Driving support becomes possible.

In the present invention, the vehicle conditions related to vehicle performance are specified using at least one of vehicle length, vehicle width, vehicle height, wheelbase, distance from the vehicle head to the driver's seat, turning radius, and vehicle type. is characterized by

Vehicle conditions include vehicle length, vehicle width, vehicle height, wheelbase, distance from the front of the vehicle to the driver's seat, turning radius, and vehicle type. Become.

The present invention is characterized in that the environmental conditions around the vehicle while it is running are identified using at least one of a calendar, time of day, weather, and illuminance.

For example, depending on the time of day, there are cases where the driving behavior is different, such as being biased toward one of a plurality of lanes. Therefore, by considering the surrounding environmental conditions when the vehicle is running, it is possible to provide more precise driving assistance.

A fourth invention , in addition to the basic configuration, further has moving object information detection means for detecting moving object information from infrastructure sensors including surveillance cameras installed along the travel path, and is detected by the moving object information detection means. The moving object information obtained is added to the information on the travel history of the vehicle passing through the travel road in the area where the infrastructure sensor is installed.

By using infrastructure sensors, it is possible to significantly increase the amount of information that constitutes the elements for creating a driving behavior database.

The driving behavior database of the present invention is used when a driving assistance device executes guidance regarding at least one of possible routes and speeds for a vehicle traveling on a road in a specific area, and At least one of the vehicle conditions related to maneuverability and the surrounding environmental conditions when the vehicle is running is used as a classification requirement, and the road area centered on the lane of the driving road is expressed as a two-dimensional grid map, and the two-dimensional grid map is: It is characterized in that the driving action data capable of representing the traveling road in the specific region is stored by the spatial distribution obtained by projecting the information related to the traveling history aggregated for each lane of the traveling road.

According to the present invention, at least one of the vehicle conditions related to the motion performance of the vehicle and the environmental conditions surrounding the vehicle while the vehicle is running is used as a classification criterion, and the road area centering on the lane of the driving road is expressed as a two-dimensional grid map. and storing, as a driving behavior database, driving behavior data capable of representing the driving road in the specific area by a spatial distribution obtained by projecting information related to driving history information aggregated for each lane of the driving road on the two-dimensional grid map. and utilize it for driving support equipment. As a result, it is possible to support automatic driving based on driving behavior close to that of a human by being used in a driving behavior support device that expresses a two-dimensional spatial distribution on a road.

As described above, according to the present invention, it is possible to generate a driving action database that takes into consideration local rules, knowledge suitable for the surrounding environment, and empirical know-how.

In addition to the effects described above, it is possible to support automatic driving based on driving behavior similar to that of humans by using it in a driving behavior support device that expresses a two-dimensional spatial distribution on a road.

1 is a schematic configuration diagram of a driving behavior database generation device according to a first embodiment; FIG. FIG. 2 shows an example of lane-level road structure data according to the first embodiment, where (A) is a plan view of an intersection where two roads intersect, and (B) is a map of surrounding road data based on input position information. is. FIG. 2 is a plan view showing a target for estimating a vehicle position according to the first embodiment; FIG. It is an environmental condition table specified by the relationship between the illumination and the wiper according to the first embodiment. It is a plan view that visualizes the route and speed distribution according to the first embodiment, (A) is a plan view around the intersection, (B) is a route and speed distribution visualized on the plan view of the intersection. It is a distribution map displayed by 4 is a control flowchart showing a database generation procedure according to the first embodiment; FIG. 2 is a schematic configuration diagram of a driving behavior database generation device according to a second embodiment; FIG. (A) to (C) are perspective views showing an installation state of a camera applied as an infrastructure sensor. FIG. 10 is a perspective view showing the positional relationship between the vehicle and the camera when estimating the vehicle position by the camera based on triangulation according to the second embodiment; It is a top view of the road which concerns on 2nd Embodiment, (A) is a top view of an intersection, (B) is a top view of a junction where the number of lanes increases/decreases. 9 is a control flow chart showing a database generation procedure according to the second embodiment;

In the present invention, data collected and detected during travel is collected from a plurality of vehicles that have traveled on the same road section, and a database for designing appropriate traffic routes and speed control is automatically generated.

The present invention will be described in detail in the following first and second embodiments. In the first and second embodiments, lane-level road network maps include intersections The explanation will focus on adding information on optimal traffic routes and optimal speeds (including stops) on specific road sections.

That is, it is assumed that a lane-level road network map has already been prepared. Also, travel data is transmitted from the vehicle to a center server (not shown).

Furthermore, in the following descriptions of the first and second embodiments, data processing on the vehicle side and data processing on the center server side are not clearly distinguished. This is because the data processing on the vehicle side varies depending on the processing power of the on-vehicle computer. Also, the arrangement of each function shown below does not affect the present invention.

"First Embodiment"

FIG. 1 is a schematic configuration diagram of a driving behavior data generation device 10 according to the first embodiment.

As shown in FIG. 1 , the driving behavior data generating device 10 includes an in-vehicle sensor group 12 , a driving information generating section 14 and a driving information generating section 16 . As a storage medium for storing information, a road structure storage unit 18 is connected to the driving information generation unit 14, and a vehicle information storage unit 20 and a driving route storage unit 22 are connected to the driving information generation unit 16. there is

(Vehicle sensor group 12)

The vehicle-mounted sensor group 12 used in the first embodiment includes an environment recognition sensor 24 , a vehicle motion sensor 26 , a vehicle state sensor 28 and a positioning sensor 30 .

The environment recognition sensor 24 is assumed to be a camera, laser radar, millimeter wave radar, ultrasonic sensor, or the like, detects moving objects and obstacles existing around the vehicle, and obtains their three-dimensional positions and velocities. It also searches for landmarks to be compared with the map to estimate the vehicle position, and detects lane marks necessary for lane estimation. Observation range and resolution are design items of the sensor, and are not particularly limited here.

The vehicle motion sensor 26 includes a gyro sensor, a wheel speed sensor, and the like, and measures vehicle motion information such as vehicle speed, yaw rate, acceleration, angular velocity, and steering angle at predetermined intervals.

A vehicle state sensor 28 acquires signals regarding the operating status of turn signals, lights, wipers, and the like. It also acquires signals relating to the operational status of safety systems such as ABS (Antilock Brake System), VSC (Vehicle Stability Control), LDW (Lane Departure Warning), and PCS (Pre-Crash Safety).

The positioning sensor 30 is mainly assumed to be a GPS (Global Positioning System), and measures the absolute position, moving direction, and speed of the vehicle based on signals received from GPS satellites. Time information can also be obtained from GPS.

(storage medium)

The road structure storage unit 18 stores lane-level road network map data. FIG. 2 shows an example of lane-level road structure data.

FIG. 2A is a plan view of an intersection X where two roads R1 and R2 intersect, and the road structure storage unit 18 has information such as the number of lanes, their center positions, lane widths, traveling directions, connection relationships, and the like. . The road structure storage unit 18 provides surrounding road data Dx, as shown in FIG. 2B, based on the input positional information. Moreover, the road structure storage unit 18 preferably has not only the road data Dx but also landmark features necessary for position estimation.

The vehicle information storage unit 20 stores information about the size and motion performance of the vehicle. It has at least a plurality of pieces of information out of vehicle length, vehicle width, vehicle height, wheelbase, turning radius, distance from the vehicle head to the driver's seat, vehicle type, and weight, and is associated with the vehicle type ID. If the necessary information is flowing through the in-vehicle network, the vehicle information storage unit 20 may not be separately prepared, and the information may be included in the travel information aggregated to a center server (not shown) and transmitted.

The driving route storage unit 22 stores finally generated route information and speed information. The route information and speed information stored in the driving route storage unit 22 (corresponding to the driving action database of the present invention) are stored in association with the lane IDs, road link IDs, and intersection IDs of the road structure storage unit 18 .

(Travel information generator 14)

Based on the detection (measurement) information received from each sensor of the vehicle-mounted sensor group 12, the driving information generation unit 14 extracts information necessary for generating a driving behavior database.

As shown in FIG. 1 , the travel information generation unit 14 has a travel lane estimation unit 32 , a vehicle state detection unit 34 , a surrounding environment recognition unit 36 , a travel route generation unit 38 , and an exceptional travel determination unit 40 . Information necessary for generating the driving behavior database is first generated for each vehicle.

The driving lane estimating unit 32 refers to the map data based on the positioning information from the GPS, compares the landmarks stored in the road structure storage unit 18 with the measurement results of the environment recognition sensor 24, and determines the location in the map data. Estimates the position of the vehicle and identifies the current lane.

Lane marks, three-dimensional objects such as signs, signals, signboards, and image feature values can be used as landmarks used for estimating the position of the vehicle. FIG. 3 shows an image of vehicle position estimation.

As shown in FIG. 3, the position of the vehicle 42 can be estimated from the directions of the lane marks 46 and 48, the U-turn prohibition sign 50, and the speed limit sign 52 visible from the vehicle 42. FIG.

As shown in FIG. 1, the vehicle state detector 34 analyzes signals from the vehicle state sensor 28 to detect their operating states.

The surrounding environment recognition unit 36 uses the environment recognition sensor 24 to detect moving objects, parked vehicles, and other obstacles that interfere with the travel path of the own vehicle.

The surrounding environment recognizing unit 36 may track moving objects such as a preceding vehicle and an oncoming vehicle in the time direction while considering the position of the own vehicle, and calculate their travel trajectories. In this case, necessary vehicle information such as the size of the vehicle is also detected and attached to the travel locus.

The travel route generator 38 calculates the travel locus of the own vehicle. Since locus information with high local accuracy is required here, a locus connecting GPS observation results cannot be used.

For example, the measurement value of the vehicle motion sensor 26 and GPS Doppler information are merged, or the vehicle position estimated by the driving lane estimation unit 32 is tracked and connected smoothly.

If a camera or Lidar (Light Detection and Ranging) is used as the environment recognition sensor 24, the motion amount of the sensor can be estimated by time-series tracking of stationary objects in the observation results, and the vehicle motion can be estimated with high accuracy. This is a technique called visual odometry.

The exceptional travel determination unit 40 aggregates the obstacle information detected by the surrounding environment recognition unit 36, the vehicle trajectory estimated by the travel route generation unit 38, and the operation status of the winkers to evaluate the vehicle motion.

If the vehicle movement is evaluated as a lane change or an avoidance action against a vehicle parked on the road, it is determined that it is not a regular driving route, and the data is not used for subsequent analysis processing. One of the judgment criteria is whether or not the travel locus deviates from the own vehicle lane.

(Driving information generator 16)

The driving information generator 16 aggregates the driving information generated for each vehicle, and generates information related to driving behavior such as the actual driving route, speed, and stop position for each lane.

The driving information generator 16 has a travel data classifier 54 , a route information generator 56 , and a speed information generator 58 .

The traveling data classification unit 54 classifies the traveling information by vehicle type and by environment.

A classification method based on vehicle conditions and environmental conditions is prepared in advance in the traveling data classification unit 54 . The route and speed change depending on the size, turning radius, and weight of the vehicle. In addition, the traffic volume changes depending on the time of day and the weather, and there is a high possibility that changes will appear in driving behavior. The classification requirements are designed so that path distributions and velocity distributions can be generated for each of these factors.

Here, in the surrounding environment recognition unit 36 (details will be described later), when the travel locus of a moving vehicle other than the own vehicle is calculated, each travel locus is similarly classified. Classification by vehicle conditions is based on vehicle length, vehicle height, vehicle width, wheelbase, distance from the front of the vehicle to the driver's seat, turning radius, and vehicle type. It is better to set several categories from the viewpoint of vehicle control rather than finely classifying them numerically.

On the other hand, environmental conditions may be obtained by accessing a website (Web) on the center server side and associating weather conditions and lighting conditions corresponding to travel data based on location and time.

As shown in FIG. 4, it is also effective to omit the search by associating the vehicle state with the weather condition in advance.

FIG. 4 is identified by the operational relationship between the vehicle lights and the wipers. When the wipers are off, the weather is fine, and intermittent operation → continuous operation of slow and steady operation is used to distinguish light rain, driving rain, and heavy rain. On the other hand, when all the lights are off, it is daytime, and the lighting states of small lights (car width lights, taillights, etc.), headlamps (headlights), and fog lamps distinguish between evening, night, and fog. In addition, since there are vehicles with all-day lighting due to automatic lights, it is necessary to change the table according to the vehicle type, model, etc.

The route information generation unit 56 extracts the travel locus information from the travel data classified into the same category, and generates the spatial distribution thereof.

For example, a road area centered on a lane is expressed as a two-dimensional grid map, and the running locus is projected onto the grid map.

Each grid is configured to store the number of times of projection, and the number of times of projection of the grid overlapping the running locus is increased. By sequentially projecting travel data classified into the same category, a two-dimensional histogram or a two-dimensional probability distribution is finally formed, and the route that the vehicle actually passes can be expressed.

Two-dimensional grid maps are useful for visualizing distributions and integrating with other information. However, in order to reduce the size of data to be stored, other representation methods can be used. For example, after examining the route distribution using a two-dimensional grid map, it is possible to extract routes with high frequencies and apply models such as splines and clothoids to parametrically represent representative routes.

Alternatively, the representative route may be expressed simply by the amount of offset from the center of the lane. The number of representative routes is not limited to one, and multiple candidates may be generated. Furthermore, while the number of times of projection does not exceed a predetermined threshold, it may be determined that reliability is low, and output may be withheld until a sufficient amount of data is accumulated.

The speed information generator 58 generates a distribution of speed information associated with the travel locus.

A two-dimensional grid map is prepared (or shared) in the same manner as the route information generator 56 described above. In the path information generator 56, each grid is made to store the number of times of projection, but here each grid is made to have a velocity distribution. For example, a memory capable of storing a one-dimensional histogram may be reserved for each grid, and the velocity distribution may be stored in histogram format. A statistic expressed by an average and a variance may be used, or only representative values such as maximum and minimum values may be stored.

In addition, the speed information generation unit 58 separately calculates the frequency of locations where the minimum value of the speed distribution is 0, that is, where the vehicle has stopped. By extracting places where the frequency exceeds a predetermined value, it is possible to automatically extract places where the vehicle should stop, such as before entering an intersection or waiting for a right turn within an intersection. Furthermore, the speed information generator 58 may spatially express the stop location on a two-dimensional grid map, or generate a virtual stop line by applying line segments to several peak positions. good too.

The route information and speed information need not be generated all at once. If traveling data is collected sequentially, the interim result of the statistical processing is temporarily stored in the driving route storage unit 22, and the interim result is obtained by accessing the driving route storage unit 22 at the next processing timing. Then, additional data processing can be performed.

Also, in the first embodiment, the configuration for processing both route information and speed information has been described, but either one may be generated. FIG. 5 is an example of visualizing the distribution of paths and velocities.

FIG. 5A is a plan view of an intersection X (within a dotted line frame) where two roads R1 and R2 intersect and its surroundings. FIG. 5(B) visualizes and displays the route and speed distribution on the plan view of the intersection X in FIG. 5(A). In FIG. 5B, belt-shaped lines L of different densities on the image of the intersection X represent the route, and the density difference of the lines L represents the vehicle speed. The higher the concentration, the faster the vehicle speed. In the case of visualizing, not only identification by density but also by color may be used.

The operation of the first embodiment will be described below with reference to the flowchart of FIG.

At step 100, the data detected by the vehicle-mounted sensor group 12 is obtained, and then the process proceeds to step 102 to estimate the lane-level vehicle position based on the vehicle-mounted sensor group 12 and the road structure information.

In the next step 104, the travel locus (with speed) of the own vehicle is calculated, then the process proceeds to step 106 to detect obstacles existing on the travel of the own vehicle, and the process proceeds to step 108. FIG. At step 108, it is determined whether or not the vehicle is running exceptionally based on the obstacle and the shape of the track.

In the next step 110, each travel data obtained in the processing of steps 100 to 108 is aggregated in the center server, and the process proceeds to step 112. FIG.

At step 112, the vehicle information and environment information related to each travel data are obtained, then the process proceeds to step 114 to classify the travel data based on the vehicle information and environment information, and the process proceeds to step 116.

At step 116, the path distribution is generated and updated for each category. Also, in the next step 118 , the speed distribution on the route is generated for each category, updated, and the process proceeds to step 120 .

At step 120, the generated distribution information (route, speed) is stored in the driving route storage unit 22, and this routine ends.

The distribution information stored in the driving route storage unit 22 is processed to represent a two-dimensional spatial distribution on the road surface, for example, for vehicles that are automatically driven by a driving behavior support device (automatic driving system). Applies when

For example, many of the automated driving systems that have been proposed in the past use lane-level road network information to generate routes to destinations, and are basically controlled to run along the center of the lane. It is designed to generate local avoidance paths when encountering obstacles.

Here, the shape of the local route and the speed control for passing through it are very important from the viewpoint of passenger's sense of security and comfort.

Local routes may be ruled based on traffic rules and driving know-how, but it is very difficult to design them so as to correspond to behavior patterns according to the situation and stopping places according to the environment.

In the first embodiment, the local driving behavior that the driver (human) customarily performs (for example, taking a course in an intersection, stopping at an intersection without a stop line, multiplexing a single lane, etc.) Usage, etc.) distribution information is aggregated and managed and stored in association with road network information.

By using this distribution information, the automated driving system can realize automated driving that is suitable for the environment and that does not cause discomfort to the occupants. This allows the automated driving system to plan human-like driving behaviors according to the environment and weather.

In addition, in the first embodiment, since the database is generated by automatic processing using the vehicle-mounted sensor group 12, manual processing by the operator is not required and information can be generated at low cost.

"Second Embodiment"

A second embodiment of the present invention will be described below. In the second embodiment, the same reference numerals are assigned to the same configurations as those of the first embodiment, and the description of the configurations is omitted.

In contrast to the first embodiment described above, the features of the second embodiment include the following features.

(Feature a) Use of information from infrastructure sensors installed near roads, such as surveillance cameras

(Feature b) Collecting information on dangerous driving and the occurrence of various system errors as exceptional driving and generating lane-level risk information.

(Feature c) Set road sections as units for aggregating data based on road structure, and change how statistics are held for each section.

FIG. 7 is a schematic configuration diagram of a driving action data generating device 10A according to the second embodiment of the invention.

As shown in FIG. 7, the driving action data generation device 10A includes an in-vehicle sensor group 12, a driving information generation section 14, and a driving information generation section 16A. As a storage medium for storing information, a road structure storage unit 18 is connected to the driving information generation unit 14, and a vehicle information storage unit 20 and a driving route storage unit 22 are connected to the driving information generation unit 16A. there is

Furthermore, infrasensor 60 is utilized in the second embodiment.

The travel information generation unit 14 has a travel lane estimation unit 32 , a vehicle state detection unit 34 , a surrounding environment recognition unit 36 , a travel route generation unit 38 , and an exceptional travel determination unit 40 .

The driving information generation unit 16A has a travel data classification unit 54, a route information generation unit 56, a speed information generation unit 58, a risk information generation unit 62, and a road section setting unit 64.

Here, in the above-described first embodiment, the configuration is such that the information of the in-vehicle sensor group 12 mounted on the vehicle is collected. On the other hand, in the second embodiment, one of the features is that the infrastructure sensor 60 is used as an information source other than the vehicle-mounted sensor group 12 .

As the infrastructure sensor 60, cameras 60A, 60B, and 60C can be applied according to their respective uses, as shown in FIGS. 8(A) to 8(C).

In other words, surveillance cameras 60A and 60B (see FIGS. 8A and 8B) are often installed around roads centering on intersections. In addition, surveillance cameras 60C (see FIG. 8C) and the like are often installed in roadside shops and buildings for the purpose of crime prevention and the like.

With the spread of IoT (Internet of Things), there is a good possibility that detection (measurement) data from the infrastructure sensors 60 represented by the cameras 60A, 60B, and 60C can be used in the future. Information from the fixed infrastructure sensor 60 is very useful for observing road conditions.

As shown in FIG. 7, the infrasensor 60 is connected to a moving object information extraction unit 68, and the image data captured by the infrasensor 60 is sent to the moving object information extraction unit together with the position/orientation information of the infrasensor 60. sent to 68.

The cameras 60A, 60B, and 60C (see FIG. 8) applied as the infrastructure sensor 60 are basically fixed and their absolute positions do not change. If the orientations (imaging optical axes) of the cameras 60A, 60B, and 60C can be controlled by panning, tilting, zooming, etc., parameters including optical axis direction information at the time of imaging can be associated with the image. preferable.

The moving object information extracting unit 68 extracts a moving vehicle from the images taken by the cameras 60A, 60B, and 60C, for example, by following the procedures (1) and (2), and tracks the position of the vehicle in the time direction. Detect trajectories.

(Procedure 1) First, moving object area candidates are detected in an image using background difference and time difference. A search area is set in the screen to include the candidate area, and the vehicle is detected using pattern recognition processing.

(Procedure 2) Assuming that the installation heights and line-of-sight directions of the cameras 60A, 60B, and 60C are known, the lane-level vehicle position is estimated using means such as the principle of triangulation (see FIG. 9).

As shown in FIG. 9, a vehicle V is positioned on the road surface, and one of the cameras 60A, 60B, and 60C (for example, the camera 60A) is positioned just above the vehicle V at a distance d along the road surface. is set up. It is assumed that the camera 60A is installed at a height h from the road surface.

The position of the vehicle V can be specified by setting the angle θ with respect to the vertical line of the camera 60A when the vehicle V is captured by the camera 60A.

If the camera 60A has a sufficient resolution, lane marks on the road surface can also be detected, so it is possible to determine the driving lane.

As shown in FIG. 7, the moving object information extraction unit 68 is connected to the travel data classification unit 54 of the driving information generation unit 16A. The trajectory information of the moving object detected in the above procedures (1) and (2) is transmitted to the driving information generator 16A, and integrated with the driving information based on the vehicle-mounted sensor group 12 for use.

It is also effective to merge (integrate) the information of the infrastructure sensor 60 with the information of the vehicle-mounted sensor group 12 by setting an appropriate weight. The weight may be changed according to the position of the imaging target. By integrating the information of the infrastructure sensor 60, it is possible to improve the accuracy and reliability of the generated information and to generate near-real-time information.

A risk information generator 62 is connected to the travel data classifier 54 .

Here, in the above-described first embodiment, a trajectory that deviates from the own vehicle lane is determined as an exceptional run, and is not used for statistical processing. On the other hand, in the second embodiment, information on exceptional driving is actively collected, and lane-level risk information is generated.

In the second embodiment, similarly to the first embodiment, avoidance of parked vehicles and lane changes are excluded from statistical processing.

On the other hand, when a sudden steering operation or acceleration/deceleration occurs from the output of the vehicle motion sensor 26, or when the operation of safety systems such as ABS, VSC, LDW, and PCS is detected from the output of the vehicle state sensor 28, It is determined as regular running, and the location of occurrence is associated with the running locus.

Information associated with the travel locus is processed by the risk information generator 62 . Like the route information generator 56, the risk information generator 62 projects the route distribution onto the two-dimensional grid map to generate the spatial distribution of event occurrence locations.

In the second embodiment, the determination result of the exceptional travel determination section 40 of the travel information generation section 14 is sent to the travel data classification section 54 via the road section setting section 64 .

The road section setting unit 64 aggregates data based on the road structure and the generated route information. This will improve the efficiency of the system.

Since the lane-level road structure is obtained from the road structure storage unit 18, road sections are set around places where the number of lanes increases or decreases, or where multiple connections occur. FIG. 10 shows a setting example of road sections. The setting of the road section may be set in advance in relation to the road structure.

In a single route where there is no increase or decrease in the number of lanes, there is little variation in the routes that can be taken, and can be generally expressed only by the amount of offset from the center position.

On the other hand, at the intersection X (see FIG. 10A) and the junction G where the number of lanes increases and decreases (see FIG. 10B), the route difference for each vehicle type appears remarkably.

Therefore, in such a region, a two-dimensional spatial distribution of paths or a three-dimensional distribution in which paths and velocities are mixed can be expressed. In this way, by changing the expression format according to the location, the required memory of the driving route storage unit 22 can be optimized.

Contrary to expectations, however, there may also occur places where the route distribution is not simple, although it is a single route. Therefore, it is desirable that the road section setting unit 64 feeds back the generated route information so that the section setting method can be changed.

The operation of the second embodiment will be described below with reference to the flowchart of FIG. It should be noted that processing steps that are the same as the predetermined procedure (see FIG. 6) in the first embodiment are appended with "A" at the end of the reference numerals.

At step 100A, the data detected by the vehicle-mounted sensor group 12 is obtained, and then the process proceeds to step 102A to estimate the lane-level vehicle position based on the vehicle-mounted sensor group 12 and the road structure information.

In the next step 104A, the travel locus (with speed) of the own vehicle is calculated, then the process proceeds to step 106A to detect obstacles existing on the travel of the own vehicle, and the process proceeds to step 108A. At step 108A, it is determined whether or not the vehicle is running exceptionally based on the obstacle and the shape of the locus, and the process proceeds to step 110A.

Here, in the infra-sensor 60, for example, acquisition of information by imaging with the cameras 60A, 60B, and 60C is executed in parallel.

In step 150, data from the infrasensor 60 is acquired, then in step 152 the moving object is tracked to generate a trajectory, and in step 154. At step 154, the information generated at step 152 is sent to the center server, and the process proceeds to step 110A.

In the next step 110A, each traveling data obtained by the processing of steps 100A to 108A is aggregated in the center server, and the process proceeds to step 111. FIG. At this time, the center server has received the information obtained from the infrastructure sensor 60 by the processing of step 154, and is aggregated in the center server together with each traveling data.

At step 111, a road section is set from the lane connection information, and the process proceeds to step 112A.

In step 112A, vehicle information and environment information related to each travel data are acquired, then the process proceeds to step 114A, the travel data is classified based on the vehicle information and environment information, and the process proceeds to step 116A.

At step 116A, a path distribution is generated and updated for each category. Also, in the next step 118A, the speed distribution on the route is generated for each category, updated, and the process proceeds to step 119A. At step 119A, the risk distribution is generated and updated for each category, and the process proceeds to step 120A.

At step 120A, the generated distribution information (route, speed) is stored in the driving route storage unit 22, and this routine ends.

According to the second embodiment, the automatic driving system can plan human-like driving behaviors according to the environment and weather.

In addition, in the second embodiment, since the database is generated by automatic processing using the vehicle-mounted sensor group 12 and the infrastructure sensor 60, manual processing by the operator is not required and information can be generated at low cost.

10 driving behavior data generation device 12 on-vehicle sensor group (acquisition means)
14 travel information generation unit (judgment means, classification means)
16 driving information generator (generating means)
18 road structure storage unit 20 vehicle information storage unit 22 driving route storage unit (driving behavior database)
24 Environment recognition sensor 26 Vehicle motion sensor 28 Vehicle state sensor 30 Positioning sensor 32 Traveling lane estimator (judgment means)
34 Vehicle state detector (acquisition means)
36 Surrounding Environment Recognition Unit (Classification Means)
38 travel route generator (judgment means)
40 Exceptional travel determination unit 42 Own vehicle 46, 48 Lane mark 50 U-turn prohibition sign 52 Speed limit sign 54 Travel data classification unit (classification means)
56 route information generator (generating means)
58 speed information generator (generating means)
R1, R2 Road X Intersection Dx Road data (Second embodiment)
10A driving behavior data generation device 16A driving information generation unit 60 infrastructure sensor 62 risk information generation unit 64 road section setting unit 60A, 60B, 60C camera 68 moving object information extraction unit

Claims (9)

Acquisition means for acquiring driving information related to driving of the vehicle from an in-vehicle sensor mounted on the vehicle;
Determining means for comparing the driving information acquired by the acquisition means with known structure information of the traveling road, and determining the traveling history including the vehicle position and the traveling trajectory for each lane of the traveling road composed of a plurality of lanes. When,
Classification means for classifying the information about the travel history determined by the determination means by lane unit of the travel road, using at least one of the vehicle conditions related to the motion performance of the vehicle and the environmental conditions around the vehicle when the vehicle is traveling as classification criteria;
For each classification requirement classified by the classification means, driving behavior data used when executing guidance regarding at least one of possible routes and speeds that the vehicle can take for a vehicle traveling on a road in a specific area. and generating means for generating
Guidance on the route,
A road area centered on the lane of the travel road is expressed as a two-dimensional grid map, and a spatial distribution is generated by projecting information related to travel history aggregated for each lane of the travel road onto the two-dimensional grid map. A driving action data generation device characterized by: Acquisition means for acquiring driving information related to driving of the vehicle from an in-vehicle sensor mounted on the vehicle;
Determining means for comparing the driving information acquired by the acquisition means with known structure information of the traveling road, and determining the traveling history including the vehicle position and the traveling trajectory for each lane of the traveling road composed of a plurality of lanes. When,
Classification means for classifying the information about the travel history determined by the determination means by lane unit of the travel road, using at least one of the vehicle conditions related to the motion performance of the vehicle and the environmental conditions around the vehicle when the vehicle is traveling as classification criteria;
For each classification requirement classified by the classification means, driving behavior data used when executing guidance regarding at least one of possible routes and speeds that the vehicle can take for a vehicle traveling on a road in a specific area. generating means for generating;
Surrounding environment recognition means for recognizing the surrounding environment including the presence or absence of facilities from the driving information acquired from the vehicle-mounted sensor;
Based on the information about the surrounding environment recognized by the surrounding environment recognition means and the running trajectory of the vehicle, whether or not there was an exceptional run including at least one of a lane change and an obstacle avoidance action with respect to the trajectory. and an exceptional travel determination means for determining
Excludes a travel trajectory determined as an exceptional travel from being classified by the classifying means;
A driving action data generation device characterized by: Acquisition means for acquiring driving information related to driving of the vehicle from an in-vehicle sensor mounted on the vehicle;
Determining means for comparing the driving information acquired by the acquisition means with known structure information of the traveling road, and determining the traveling history including the vehicle position and the traveling trajectory for each lane of the traveling road composed of a plurality of lanes. When,
Classification means for classifying the information about the travel history determined by the determination means by lane unit of the travel road, using at least one of the vehicle conditions related to the motion performance of the vehicle and the environmental conditions around the vehicle when the vehicle is traveling as classification criteria;
For each classification requirement classified by the classification means, driving behavior data used when executing guidance regarding at least one of possible routes and speeds that the vehicle can take for a vehicle traveling on a road in a specific area. generating means for generating;
An irregular driving extracting means for extracting irregular driving based on statistical data of driving operations including sudden braking and sudden steering based on the driving information and the operation status of the safety system installed in the vehicle,
The running trajectory extracted as the irregular running is registered as risk information for classification requirements by the classification means.
A driving action data generation device characterized by: Acquisition means for acquiring driving information related to driving of the vehicle from an in-vehicle sensor mounted on the vehicle;
Determining means for comparing the driving information acquired by the acquisition means with known structure information of the traveling road, and determining the traveling history including the vehicle position and the traveling trajectory for each lane of the traveling road composed of a plurality of lanes. When,
Classification means for classifying the information about the travel history determined by the determination means by lane unit of the travel road, using at least one of the vehicle conditions related to the motion performance of the vehicle and the environmental conditions around the vehicle when the vehicle is traveling as classification criteria;
For each classification requirement classified by the classification means, driving behavior data used when executing guidance regarding at least one of possible routes and speeds that the vehicle can take for a vehicle traveling on a road in a specific area. generating means for generating;
moving object information detection means for detecting moving object information from infrastructure sensors including surveillance cameras installed along the travel path;
adding the moving object information detected by the moving object information detecting means to information related to the travel history of the vehicle passing through the travel route in the area where the infrastructure sensor is installed;
A driving action data generation device characterized by: Guidance on the speed,
5. The driving action data generation device according to claim 1, wherein the generation is a vehicle speed distribution along the running road of the specific region . further comprising section setting means for setting road sections for classifying the information related to the travel history determined by the determination means;
The section setting means sets the sections of the traveling road based on points of change including points at which the number of lanes on the traveling road increases and decreases, and points at which the traveling road merges and diverges. 1. The driving behavior data generation device according to any one of the items. The vehicle condition related to the performance of the vehicle is
It is specified using at least one of vehicle length, vehicle width, vehicle height, wheel base, distance from the vehicle head to the driver's seat, turning radius, and vehicle type. 1. The driving behavior data generating device according to 1. Surrounding environmental conditions when the vehicle is running are
8. The driving behavior data generating device according to any one of claims 1 to 7, characterized by using at least one of calendar, time period, weather, and illuminance. The driving support device is used when executing guidance on at least one of the route or speed that the vehicle can take for a vehicle traveling on a road in a specific area,
Using at least one of the vehicle conditions related to the motion performance of the vehicle and the surrounding environmental conditions when the vehicle is running as a classification requirement, the road area centering on the lane of the driving road is expressed as a two-dimensional grid map, and the two-dimensional grid map (2) a driving behavior database storing driving behavior data capable of expressing the driving road in the specific area by a spatial distribution obtained by projecting information related to driving history aggregated in units of lanes of the driving road;

Images