JP2021009655A - Surrounding object recognition method and surrounding object recognition device - Google Patents

Abstract

To execute a driving support suitable for a driving situation without distinguishing and recognizing each object in a scene where there is an object that blocks a planned driving lane of an own vehicle.SOLUTION: A surrounding object recognition device comprises: a rider 113 that acquires point cloud data Pn of surrounding environment; and a surrounding object recognition unit 315 that recognizes an object existing around an own vehicle V based on the point cloud data Pn acquired by the rider 113. In a surrounding object recognition method, the surrounding object recognition unit 315 recognizes a lane blockage region LSA, which is a region in which a left lane LL is blocked by parked vehicles PV1 and PV2, which hinder a traveling of the own vehicle V, based on the point cloud data Pn. Then, the lane blockage region LSA is expanded by connecting the point cloud data Pn in the lane traveling direction starting from a starting point SP of the lane blockage region LSA.SELECTED DRAWING: Figure 8

Description

The present disclosure relates to a peripheral object recognition method and a peripheral object recognition device for recognizing an object existing around the own vehicle.

Conventionally, the event information acquisition unit receives information on event information sent from the analysis unit and the communication unit that may interfere with the traveling of a moving object, and from the information, the presence / absence of a vehicle parked on the street, the presence / absence of a traffic jam, and the presence / absence of a traffic jam. Derived event information including the traveling obstruction section. When the vehicle is parked on the road on the traveling lane and the traveling lane is the traveling lane, the determination unit determines that the first lane change is necessary before the moving body reaches the traveling obstruction section. Then, when the section from the beginning of the traveling obstruction section on the side close to the guidance attention point to the guidance attention point is not a lane change prohibition section, the determination unit determines the first lane after passing through the travel obstruction section. It is determined that the lane change of 2 is necessary. Subsequently, an information generation device and an information generation method are disclosed in which the generation unit generates guidance information of the content of the determination result by the determination unit.

Japanese Unexamined Patent Publication No. 2003-159966

In the prior art example disclosed in Patent Document 1, a parked vehicle is recognized by recognizing a hazard state of the vehicle by using a camera. However, the second and subsequent vehicles in the parked convoy may be blind spots for the first vehicle, and it is often difficult to recognize them as vehicles. In that case, an inappropriate driving lane plan is made for an object behind (the direction of lane travel) following the recognized object, such as a second parked vehicle object, and driving support suitable for the driving situation is executed. There was a problem that it could not be done.

This disclosure focuses on the above-mentioned problems, and provides driving support suitable for a driving situation without distinguishing and recognizing each object in a scene in which an object blocking the planned driving lane of the own vehicle exists. The purpose is to execute.

In order to achieve the above object, the present disclosure includes a sensor that acquires spatial position information of an object in the surrounding environment and a controller that recognizes an object existing around the own vehicle by the spatial position information acquired by the sensor. Be prepared. The controller recognizes the lane blockage area by the following procedure. Based on the spatial position information, the lane obstruction area, which is the area where the lane is obstructed by an object that hinders the traveling of the own vehicle, is recognized. The lane obstruction area is expanded by connecting the spatial position information of the object in the lane traveling direction starting from the start point of the lane obstruction area.

Since the above-mentioned problem-solving means is adopted, in a scene in which an object blocking the planned traveling lane of the own vehicle exists, it is possible to execute the traveling support suitable for the traveling situation without distinguishing and recognizing each object.

FIG. 5 is an overall system configuration diagram showing a system configuration of an autonomous driving vehicle to which the surrounding object recognition method and the surrounding object recognition device of the first embodiment are applied. It is a block block diagram which shows the control block structure of an automatic operation controller. It is a block block diagram which shows the structure of the surrounding object recognition part of the automatic driving controller. It is a flowchart which shows the flow of the surrounding object recognition control processing executed in the surrounding object recognition part. It is a block block diagram which shows the structure which recognizes the surrounding object of the own vehicle by the background technology. It is a problem explanatory diagram which shows the problem in the background technology. It is an action explanatory view which shows the start point recognition action of the lane blockage region and the expansion action of a lane blockage region in this disclosure. It is an effect explanatory diagram which shows the target locus generation using the extended lane blockage area information recognized by the surrounding object recognition method in this disclosure. It is an action explanatory diagram which shows the recognition action of a lane blockage candidate by the point cloud data acquired from a rider. It is an action explanatory diagram which shows the recognition action of a parked vehicle and the start point setting action of a lane blockage area by the point cloud data acquired from a rider. It is an action explanatory view which shows the lane blockage area expansion action which expands a lane blockage area by a vertical deviation distance condition by a point cloud data acquired from a rider. It is an action explanatory diagram which shows the example which excludes from the connection target by the lateral deviation distance condition in the expansion action of the lane blockage area by the point cloud data acquired from a rider. It is an action explanatory diagram which shows the example which excludes from the connection target by the judgment that it is a moving object in the expansion action of the lane blockage area by the point cloud data acquired from a rider. It is an operation explanatory view which shows the expansion processing action of a lane blockage area when the start point of a lane blockage area cannot be recognized.

Hereinafter, a mode for implementing the surrounding object recognition method and the surrounding object recognition device according to the present disclosure will be described based on the first embodiment shown in the drawings.

In the surrounding object recognition method and the surrounding object recognition device in the first embodiment, a target locus is generated when the automatic driving mode is selected, and the vehicle motion is controlled by the speed and the steering angle so as to travel along the generated target locus. It is applied to an autonomous driving vehicle (an example of a driving support vehicle). Hereinafter, the configuration of the first embodiment will be described separately as "overall system configuration", "control block configuration of the automatic operation controller", "detailed configuration of the surrounding object recognition unit", and "surrounding object recognition control processing configuration".

[Overall system configuration (Fig. 1)]
As shown in FIG. 1, the autonomous driving vehicle AD includes an in-vehicle sensor 1, a navigation device 2, an in-vehicle control unit 3, an actuator 4, and an HMI module 5. "HMI" is an abbreviation for "Human Machine Interface".

The in-vehicle sensor 1 is various sensors mounted on the own vehicle in order to recognize an object around the own vehicle, a surrounding environment such as a road shape, and a state of the own vehicle. The in-vehicle sensor 1 has an external sensor 11, a GPS receiver 12, and an internal sensor 13. In the in-vehicle sensor 1, sensor fusion may be performed to acquire necessary information by using a plurality of different sensors.

The external sensor 11 is a detection device that detects environmental information around the vehicle. The external sensor 11 includes, for example, a camera, a radar (“Radar”: an abbreviation for Radio Detection and Ranging), a lidar (“Lidar”: an abbreviation for Light Detection and Ranging), and a sonar (“Sonar”: an abbreviation for Sound Navigation and Ranging). ) Etc.

The GPS receiver 12 is a device that receives signals from three or more GPS satellites by the GNSS antenna 12a and acquires position data (latitude and longitude) indicating the position of the own vehicle. In addition, "GNSS" is an abbreviation for "Global Navigation Satellite System", and "GPS" is an abbreviation for "Global Positioning System". Further, when the signal reception by the GPS receiver 12 is poor, the function of the GPS receiver 12 may be complemented by using the internal sensor 13 or the odometer (vehicle movement amount measuring device).

The internal sensor 13 is a detection device that detects own vehicle information such as speed, acceleration, and attitude data of the own vehicle. For example, it has a 6-axis inertial measurement unit (IMU) and can detect the moving direction, direction, and rotation of the own vehicle. Further, the moving distance, the moving speed, and the like can be calculated based on the detection result of the internal sensor 13. The 6-axis inertial sensor is realized by combining an acceleration sensor capable of detecting acceleration in three directions of front-back, left-right, and up-down, and a gyro sensor capable of detecting the speed of rotation in these three directions. The internal sensor 13 can include necessary sensors such as a wheel speed sensor, a yaw rate sensor, and an accelerator operation amount sensor.

Further, in the in-vehicle sensor 1, necessary information may be acquired from the outside by performing wireless communication with an external data communication device (not shown). That is, when the external data communication device is, for example, a data communication device mounted on another vehicle, vehicle-to-vehicle communication is performed between the own vehicle and the other vehicle. Through this vehicle-to-vehicle communication, necessary information can be obtained from various information held by other vehicles. Further, when the external data communication device is, for example, a data communication device provided in the infrastructure equipment, infrastructure communication is performed between the own vehicle and the infrastructure equipment. Through this infrastructure communication, necessary information can be obtained from the information held by the infrastructure equipment. As a result, for example, it is possible to supplement the map data necessary when the high-precision map data possessed by the automatic driving controller 31 has insufficient information or changed information. It is also possible to acquire traffic information such as traffic congestion information and driving regulation information on the route on which the vehicle is scheduled to travel.

The navigation device 2 is a device that incorporates facility information data and guides the route on which the vehicle travels to the destination. The navigation device 2 may include high-precision map data including position information of lanes on the road. In the navigation device 2, when the destination is input, a guide route from the current location (or an arbitrarily set departure point) of the own vehicle to the destination is generated. The generated guidance route information is combined with the high-precision map data and displayed on the display of the HMI module 5. As the destination, the one set by the occupant of the vehicle in the vehicle may be used, or the destination set by the user by the user terminal (for example, a mobile phone or a smartphone) may be set by the user via wireless communication. You may use the received destination. Further, the guide route may be calculated by a navigation device using a controller provided in the own vehicle, or may be calculated by a navigation device using a controller outside the vehicle.

The vehicle-mounted control unit 3 includes a CPU and a memory, and receives various detection information detected by the vehicle-mounted sensor 1, guidance route information generated by the navigation device 2, and driver input information appropriately input as needed. Integrated processing. The in-vehicle control unit 3 is a controller that controls vehicle motion by hierarchical processing. Note that "hierarchical processing" is to calculate the final output information by sequentially (hierarchically) executing a plurality of processes on the input information, and the output value output in the upper layer process. (Calculated value) is the input value in the lower layer processing.

The in-vehicle control unit 3 includes an automatic driving controller 31 that generates a target trajectory including a target vehicle speed profile, and a vehicle motion controller 32 that calculates a command value for controlling vehicle motion based on the generated target trajectory. Have. Here, the automatic driving controller 31 that performs the calculation in the first control cycle performs the processing of the upper layer, and the vehicle motion controller 32 that performs the calculation in the second control cycle shorter than the first control cycle performs the processing of the lower layer. Do.

The automatic driving controller 31 generates a target trajectory including a target vehicle speed profile by hierarchical processing based on input information from the vehicle-mounted sensor 1 and the navigation device 2, high-precision map data, and the like. Here, the "target locus" is a travel locus that is a target when the vehicle is automatically driven, and is, for example, a locus in which the vehicle travels within the lane width or a travelable area around the vehicle. Includes trajectories for traveling on the road and trajectories for avoiding obstacles. The information of the target trajectory including the generated target vehicle speed profile is output to the vehicle motion controller 32. The generated target trajectory information may be combined with the high-precision map data and displayed on the display of the HMI module 5.

The vehicle motion controller 32 calculates a control command value for driving the own vehicle based on the target trajectory information including the target vehicle speed profile and the input information by the driver operation (hereinafter, referred to as "driver input"). The calculated control command value is output to the actuator 4. The vehicle motion controller 32 arbitrates the traveling mode depending on the presence or absence of driver input, and calculates a control command value according to the arbitration result.

The actuator 4 is a control actuator for traveling straight / turning / stopping the own vehicle, and includes a speed control actuator 41 and a steering control actuator 42. The running includes acceleration running / constant speed running / deceleration running.

The speed control actuator 41 controls the drive torque or braking torque output to the drive wheels based on the speed control command value input from the vehicle-mounted control unit 3. As the speed control actuator 41, for example, an engine is used in the case of an engine vehicle, an engine and a motor / generator are used in the case of a hybrid vehicle, and a motor / generator is used in the case of an electric vehicle. Further, as the actuator that controls only the braking torque, for example, a hydraulic booster, an electric booster, a brake motor actuator, or the like is used.

The steering control actuator 42 controls the steering angle of the steering wheels based on the steering control command value input from the vehicle-mounted control unit 3. As the steering control actuator 42, a steering motor or the like provided in the steering force transmission system of the steering system is used.

The HMI module 5 is an interface for transmitting mutual intentions and information between the occupant including the driver of the own vehicle and the vehicle-mounted control unit 3. The HMI module 5 includes, for example, a head-up display or meter display that displays image information based on the automatic driving control status to the occupant, a speaker that outputs an announcement sound, a lamp that warns the occupant by lighting or blinking, and an operation button that the occupant performs an input operation. And touch panel.

[Control block configuration of automatic driving controller (Fig. 2)]
As shown in FIG. 2, the automatic driving controller 31 includes a high-precision map data storage unit 311, a self-position estimation unit 312, a driving environment recognition unit 313, and surroundings as information acquisition processing units necessary for generating a target trajectory. It includes an object recognition unit 315. Then, as a hierarchical processing unit for generating a target locus, a traveling lane planning unit 316, an operation determining unit 317, a traveling area setting unit 318, and a target locus generating unit 319 are provided.

The high-precision map data storage unit 311 stores high-precision three-dimensional map data (hereinafter referred to as "HD map") in which three-dimensional position information (longitude, latitude, height) of a stationary object existing outside the vehicle is set. It is an in-vehicle memory. Still objects of high-precision map data include, for example, pedestrian crossings, stop lines, various signs, branch points, road markings, traffic lights, utility poles, buildings, signs, center lines of roads and lanes, lane markings, shoulder lines, and roads. It includes various factors such as road connections. The high-precision map data storage unit 311 does not necessarily have to include all the elements of the above-mentioned stationary object.

The self-position estimation unit 312 estimates the current location (self-position) of the own vehicle on the high-precision map based on the input information. Sensor information from the vehicle-mounted sensor 1 and HD map information from the high-precision map data storage unit 311 are input to the self-position estimation unit 312. Then, the self-position estimation unit 312 estimates the self-position by matching the input sensor information with the HD map information, for example. The self-position estimation unit 312 outputs the self-position information to the driving environment recognition unit 313.

The traveling environment recognition unit 313 recognizes the traveling environment of the own vehicle by using the input information and the dynamic information that changes every moment of the own vehicle traveling environment. The driving environment recognition unit 313 contains sensor information from the in-vehicle sensor 1, guidance route information from the navigation device 2, HD map information from the high-precision map data storage unit 311, and self-position from the self-position estimation unit 312. Information etc. are entered. Then, the traveling environment recognition unit 313 calculates the traveling environment recognition information of the area around the own vehicle with reference to the estimated current location of the own vehicle. The driving environment recognition unit 313 outputs the driving environment recognition information to the operation determination unit 317.

Here, the "dynamic information" refers to, for example, information in which quasi-static data, quasi-dynamic data, and dynamic data are combined. The quasi-static data refers to data including, for example, traffic regulation information, road construction information, wide area weather information, and the like. The quasi-dynamic data refers to data including, for example, accident information, traffic congestion information, narrow area weather information, and the like. The dynamic data refers to data including, for example, peripheral vehicle information, pedestrian information, signal information, and the like. These dynamic information are layered, and the update frequency of each data is different, such that the more dynamic data changes, the higher the update frequency.

The surrounding object recognition unit 315 detects or predicts the position, attribute, and behavior of objects existing around the vehicle. Further, the surrounding object recognition unit 315 recognizes one lane blockage area by point cloud connection as a lane area to be excluded from the traveling area of the own vehicle based on the object recognition existing around the own vehicle. In the surrounding object recognition unit 315, sensor information and the like from the vehicle-mounted sensor 1 are input. Then, when the surrounding object recognition unit 315 recognizes the start point of the lane blockage region, the point cloud data from the start point is connected in the lane traveling direction to expand the lane blockage region. The surrounding object recognition unit 315 outputs lane blockage area information to the traveling area setting unit 318.

The traveling lane planning unit 316 plans the traveling lane (hereinafter, referred to as "target lane") in which the vehicle should travel on the guidance route to the destination. Guidance route information from the navigation device 2 and HD map information from the high-precision map data storage unit 311 are input to the traveling lane planning unit 316. Then, the traveling lane planning unit 316 plans the target lane from the direction of the destination determined from the route guidance information and the HD map. The target lane information is output from the traveling lane planning unit 316 to the operation determination unit 317 of the next layer.

The motion determination unit 317 extracts the events that the vehicle encounters when traveling along the target lane, and determines the "motion of the vehicle" for those events. The driving environment recognition information from the traveling environment recognition unit 313, the target lane information from the traveling lane planning unit 316, and the like are input to the operation determination unit 317. Then, the motion determination unit 317 collates the target lane with the traveling environment around the vehicle and determines an appropriate motion of the vehicle. Here, the "movement of the own vehicle" refers to the movement of the own vehicle required to drive along the target lane such as starting, stopping, accelerating, decelerating, and turning left or right. The own vehicle operation determination information is output from the operation determination unit 317 to the traveling area setting unit 318 of the next layer.

The traveling area setting unit 318 sets a traveling area in which the own vehicle can be driven along the target lane by the determined own vehicle operation. HD map information from the high-precision map data storage unit 311, own vehicle operation determination information from the operation determination unit 317, surrounding object recognition information from the surrounding object recognition unit 315, and the like are input to the traveling area setting unit 318. To. Then, the traveling area setting unit 318 collates the operation information of the own vehicle with the surrounding environment information of the own vehicle, and sets an area in which the own vehicle can travel. For example, when an object such as a parked convoy or a construction section exists around the own vehicle, a travelable area is set so as to avoid interference or contact with the area. The travelable area information is output from the travel area setting unit 318 to the target trajectory generation unit 319 of the next layer.

The target locus generation unit 319 generates a target locus of the own vehicle within the set travelable area. The travelable area information from the travel area setting unit 318 is input to the target locus generation unit 319. Then, the target locus generation unit 319 sets a constraint condition that the vehicle travels within the travelable area from the current position of the own vehicle to an arbitrarily set target position, and targets the target including the lane change by a geometric method. Generate a trajectory. The target locus generation unit 319 may generate a target locus using, for example, a composite clothoid curve. In addition, the target locus generation unit 319 may generate a target locus capable of traveling that satisfies criteria such as safety, legal compliance, and travel efficiency. The target trajectory generation unit 319 outputs the target trajectory information to the vehicle motion controller 32.

[Detailed configuration of surrounding object recognition unit (Fig. 3)]
As shown in FIG. 3, the peripheral object recognition unit 315 (controller) includes an object recognition unit 315a, a lane blockage area start point recognition unit 315b, and a lane blockage area expansion unit 315c. Then, as an in-vehicle sensor for acquiring necessary information in the surrounding object recognition unit 315, it has a camera 111, a radar 112, and a rider 113 (sensor) provided in the external sensor 11.

The camera 111 is an image pickup sensor for acquiring image data around the own vehicle. This camera is configured by combining, for example, a front recognition camera, a rear recognition camera, a right recognition camera, a left recognition camera, etc., and analyzes captured images and videos in real time using artificial intelligence and an image processing processor. Do it with. As a result, in the camera 111, objects on the own vehicle's road, lanes, objects outside the own vehicle's road (road structure, preceding vehicle, following vehicle, oncoming vehicle, surrounding vehicles, pedestrians, bicycles, two-wheeled vehicles), own vehicle's driving path (White lines on roads, road boundaries, stop lines, pedestrian crossings) ・ Road signs (speed limit) can be detected. Although it is generally not possible to measure the distance to an object with a monocular camera, it is possible to measure the distance to an object by simultaneously shooting from different viewpoints using a compound eye camera.

The radar 112 is a distance measuring sensor that receives the reflected signal of the emitted radio wave and acquires information on the direction and distance of the object. Therefore, the radar 112 can detect the positions of objects on the vehicle's road and objects outside the vehicle's road (road structures, preceding vehicles, following vehicles, oncoming vehicles, surrounding vehicles, pedestrians, bicycles, motorcycles), etc. The distance to each object can be detected.

The rider 113 measures the scattered light for the laser irradiation that emits in a pulse shape, and provides point group data (point group data) for information on the direction and distance of an object on the vehicle traveling road and an object outside the vehicle traveling road that are farther than the radar 112. It is a distance measuring sensor acquired as (spatial position information of an object). Here, the rider 113 replaces the radio waves of the radar 112 with light, emits a pulse of light outside the visible spectrum using a laser, and measures the time for the pulse to return. Then, the direction and distance where a specific pulse is reflected are recorded as point cloud data in a 3D environment centered on the rider unit. Since the rider 113 can transmit and receive pulse scans exceeding 1 million times per second, the point cloud data from the laser pulse is fine as it is called a point cloud, and the 3D environment and its contents. It is also possible to delineate the object of.

The object recognition unit 315a performs recognition processing of an object existing on the traveling path of the own vehicle based on the data acquired from the camera 111, the radar 112, and the rider 113. In this object recognition process, the object is not a moving object that does not hinder the running of the own vehicle (for example, a traveling vehicle or a paused vehicle), but a stationary object that does not hinder the running of the own vehicle (for example, a parked vehicle or construction work). Performs recognition processing as a construction sign indicating a section, a pylon, etc.).

The lane blockage region start point recognition unit 315b recognizes the start point of the lane blockage region, which is a region where the lane is blocked by a stationary object that hinders the traveling of the own vehicle. In this case, the first start point recognition pattern for recognizing the start point of the lane blockage region based on the point cloud data acquired by the rider 113 and the start point of the lane blockage region based on the object recognition result from the object recognition unit 315a. Is used in combination with the second start point recognition pattern that recognizes.

In the first start point recognition pattern, a plurality of bin division sections are set by dividing the lane in the lane traveling direction at regular intervals, and obstacles are set in the lane width direction based on the collected point group data among the plurality of bin division sections. A bin division section having a predetermined blockage width is recognized as a lane blockage candidate. Subsequently, among the point cloud data existing in the lane blockage candidates, the data point at the position closest to the target trajectory of the own vehicle is selected, and the point of the selected data point is recognized as the start point of the lane blockage region. Here, the point cloud data existing in the lane width direction from the start point of the lane obstruction region is connected to set the rear end face region, and the amount of overlap between the set rear end face region and the lane boundary on the side far from the own vehicle is When it is equal to or more than a predetermined threshold value, the recognized start point is set as the "start point of the lane blockage area by the parked vehicle".

In the second start point recognition pattern, if the object recognized by the object recognition process from the object recognition unit 315a is a stationary object that hinders the running of the own vehicle, the target locus of the own vehicle in the rear end surface region of the stationary object. The position closest to is recognized as the starting point of the lane blockage area. Here, the stationary object means a parked vehicle, a construction section, or the like.

The lane blockage area expansion unit 315c expands the lane blockage area by connecting the point cloud data acquired by the rider 113 in the lane traveling direction starting from the start point of the lane blockage area. When expanding the lane blockage area, if the point cloud data can be acquired by the rider 113, the expansion process is performed based on the acquired point cloud data, and individual object recognition is performed when the lane blockage area expansion process is executed. do not need.

When the lane blockage area is expanded starting from the start point of the lane blockage area, the vertical dissociation distance at which the two data points to be connected deviate in the lane travel direction is calculated. Then, when the vertical deviation distance condition that the vertical deviation distance is less than a predetermined threshold value is satisfied, the two data points are connected in the lane traveling direction to expand the lane blockage region. Further, if the longitudinal deviation distance is equal to or greater than a predetermined threshold value and the longitudinal deviation distance condition is not satisfied, the lane blockage area is not expanded by connecting the two data points. The "predetermined threshold" is, for example, a value equivalent to the inter-vehicle distance that can draw a target trajectory due to a lane change between two parked vehicles based on a traveling lane plan without causing contact or interference with the parked vehicle. Set. That is, for example, in the case of a parked vehicle line in which parked vehicles are connected, the vertical deviation distance condition that the vertical deviation distance is less than a predetermined threshold value is satisfied, and the expansion of the lane blockage area is allowed. When a plurality of parked vehicle rows are separated from each other, the vertical deviation distance becomes equal to or more than a predetermined threshold value, and the lane blockage area is set for each of the separated parked vehicle rows.

Here, when expanding the lane obstruction area starting from the start point of the lane obstruction area, the lateral divergence distance is calculated in which the two data points to be connected are in the lane width direction and deviate toward the side approaching the traveling lane of the own vehicle. To do. Then, when the lateral deviation distance is equal to or more than a predetermined value, the point cloud data is excluded from the connection target regardless of whether the vertical deviation distance condition is satisfied or not. Further, when the lane blockage area is expanded starting from the start point of the lane blockage area, the object attribute obtained from the data point to be connected next is determined. Then, when the object attribute is determined to be a stationary object (parked vehicle, construction section, etc.), the point group data is linked, and when the object attribute is determined to be a moving object (paused vehicle), the point group data is linked. exclude. Furthermore, when expanding the lane blockage area in a scene where the start point of the lane blockage area cannot be recognized, if there is a lane blockage area recognized in the past, a point in the lane traveling direction is based on the past lane blockage area. The lane blockage area that connects the group data is expanded.

When the lane blockage area expansion unit 315c generates lane blockage area information as surrounding object recognition information, the lane blockage area expansion unit 315c outputs the lane blockage area information to the traveling area setting unit 318. In the traveling area setting unit 318, for example, when lane blockage area information is input due to the existence of a parked vehicle line or the like or a construction section around the own vehicle, interference or contact with the lane blockage area is avoided. The travelable area is set. In the next target locus generation unit 319, the travel area setting unit 318 generates a target locus of the own vehicle within the set travelable area. That is, when the lane change is requested based on the traveling lane plan, the target locus is generated in the travelable area excluding the lane blockage area.

[Around object recognition control processing configuration (Fig. 4)]
In step S1, following the start, point cloud data is acquired from the rider 113, and the process proceeds to step S2.

In step S2, following the acquisition of the point group data in S1, of the plurality of bin division sections in which the lane is divided in the lane traveling direction, the bin division section having the blockage width in the lane width direction by the aggregated point group data. Is recognized as a lane blockage candidate, and the process proceeds to step S3.

In step S3, following the recognition of the lane blockage candidate in S2, it is determined whether or not the cause of the lane blockage is the rear end of the parked vehicle. If YES (the lane obstruction factor is the rear end of the parked vehicle), the process proceeds to step S9, and if NO (the lane obstruction factor is not the rear end of the parked vehicle), the process proceeds to step S4.

In step S4, following the determination that the lane obstruction factor in S3 is not the rear end of the parked vehicle, it is determined whether or not the lane obstruction factor is the construction section. If YES (the lane blockage factor is the construction section), the process proceeds to step S9, and if NO (the lane blockage factor is not the construction section), the process proceeds to the end.

In step S5, following the start, based on the input information acquired by the camera 111, the radar 112, and the rider 113, an object recognition process for recognizing a stationary object existing on the traveling path of the own vehicle is performed, and the process proceeds to step S6.

In step S6, following the object recognition in S5, it is determined whether or not the recognized stationary object is a stationary vehicle that hinders the traveling of the own vehicle. If YES (is a stationary vehicle), the process proceeds to step S7, and if NO (not a stationary vehicle), the process proceeds to step S8.

In step S7, following the determination in S6 that the vehicle is a stationary vehicle, it is determined whether or not the vehicle is a parked vehicle. If YES (parked vehicle), the process proceeds to step S9, and if NO (not a parked vehicle), the process proceeds to step S8.

In step S8, following the determination in S6 that the vehicle is not a stationary vehicle or the determination in S7 that the vehicle is not a parked vehicle, it is determined whether or not the recognized object is a construction sign or a pylon indicating a construction section. .. If YES (it is a construction section), the process proceeds to step S9, and if NO (not a construction section), the process proceeds to the end.

In step S9, following the determination of YES in S3 or S4 or S7 or S8, the start point of the lane obstruction region is recognized, and the process proceeds to step S10.

Here, in the first start point recognition pattern in which the lane obstruction factor is the rear end of the parked vehicle or the lane obstruction factor is the construction section, among the point group data existing in the lane obstruction candidates, the target trajectory of the own vehicle is the most. Select a data point at a close position and recognize the point of the selected data point as the starting point of the lane blockage area. Next, the point cloud data existing in the lane width direction from the start point of the lane blockage region is connected to set the rear end face region, and the amount of overlap between the set rear end face region and the lane boundary on the side far from the own vehicle is When it is equal to or more than a predetermined threshold value, the recognized start point is set as the "start point of the lane blockage area by the parked vehicle". On the other hand, in the second start point recognition pattern in which the object recognition result is the parked vehicle or the construction section, the position closest to the target locus of the own vehicle in the rear end surface region of the object indicating the parked vehicle or the construction section is the lane blockage region. Recognize as the starting point.

In step S10, following the recognition of the start point of the lane blockage area in S9 or the expansion of the lane blockage area in S11, whether or not there is a point that can be connected to the point cloud data acquired by the rider 113. to decide. If YES (there is a connectable point in the point cloud data), the process proceeds to step S11, and if NO (there is no connectable point in the point cloud data), the process proceeds to the end.

In step S11, following the determination that there is a connectable point in the point cloud data in S10, the lane blockage area expansion process by connecting the point cloud data is executed, and the process returns to step S10. That is, as long as there are connectable points in the point cloud data, the execution of the lane blockage area expansion process by connecting the point cloud data is continued.

Next, "problems of background technology and measures to solve the problems" will be described. Then, the action of the first embodiment will be described separately by dividing it into "a lane blockage region start point recognition action" and "a lane blockage region expansion action by connecting point clouds".

[Issues of background technology and measures to solve the problems (Figs. 5 and 8)]
As a background technique for setting a lane blockage region, for example, as shown in FIG. 5, the background technology performs recognition processing of an object existing on the traveling path of the own vehicle based on data acquired from a camera, a radar, or a rider. In this object recognition process, the image data from the camera is collated with the distance measurement data from the radar and rider, and stationary objects such as construction signs and pylon indicating parked vehicles and construction sections that hinder the running of the own vehicle are used as hazards. recognize. Subsequently, when the stationary object is recognized, the outer contour line of the stationary object is expanded by the interference avoidance width with the surrounding object, and the area surrounded by the enlarged contour line is recognized as the lane blockage area. Subsequently, it is common to set an area excluding the recognized lane obstruction area as a travelable area and make a travel lane plan within the travelable area.

In the above background technology, as shown in FIG. 6, the target trajectory for the own vehicle V traveling in the right lane LR to change lanes to the left lane LL in which two parked vehicles PV1 and PV2 form a parked lane. Assume an example of a scene that generates Tt. In this case, the lane obstruction region LSA is set for the first parked vehicle PV1 in the parked convoy based on the object recognition from the own vehicle V. However, vehicles after the second parked vehicle PV2 in the parked convoy are blind spots from the own vehicle V, and it is often difficult to recognize them as parked vehicles PV2, so the lane blockage area LSA is not set. It is difficult to correctly recognize an object that exists in the blind spot of an object, such as the parked vehicle PV2, because at least a part of the object recognition is hidden by the blind spot. In that case, there is a problem that the target trajectory Tt that does not consider the existence of the parked vehicle PV2 hidden in the blind spot from the own vehicle V is generated.

In response to the above problem, for obstacles in the back where object recognition is difficult due to blind spots from the own vehicle, such as the second and subsequent parked vehicles, even if object recognition is not a necessary condition, the back is in the lane blockage area. We focused on the fact that problems can be solved if obstacles can be included. Then, as a result of verifying the solution method based on the above points of interest, the present inventors can acquire the spatial position information of the object that becomes a blind spot from the own vehicle even if the entire shape of the object cannot be recognized. It was found that it is possible to expand the lane blockage area by using the spatial position information of the object.

Based on the above findings, the present disclosure is based on the rider 113 that acquires the point cloud data Pn (an example of spatial position information of an object) of the surrounding environment and the point cloud data Pn acquired by the rider 113 to surround the vehicle V. A peripheral object recognition unit 315 that recognizes an existing object is provided. The surrounding object recognition unit 315 recognizes the lane obstruction region LSA by the following procedure. The point cloud data Pn recognizes the lane obstruction region LSA, which is the region where the lane is obstructed by an object that hinders the traveling of the own vehicle V. Then, a problem-solving means was adopted in which the lane obstruction region LSA was expanded by connecting the point cloud data Pn in the lane traveling direction starting from the start point SP of the lane obstruction region LSA.

That is, as shown in FIG. 7A, the area where the left lane LL is blocked by the objects (parked vehicles PV1, PV2) that hinder the traveling of the own vehicle V according to the point cloud data (P1, ..., P9). Recognize the starting point SP (= data point P3) of the lane blockage area LSA. Subsequently, as shown in FIG. 7 (b), the lane blockage region LSA is expanded by sequentially connecting the point cloud data (P4, ..., P9) in the lane traveling direction starting from the start point SP of the lane blockage region LSA. To do. Then, as shown in FIG. 7 (c), point group data (P4, ..., P9) acquired by the rider 113 for obstacles in the back, such as the second parked vehicle PV2, which is a blind spot and whose object recognition is difficult. Is recognized as one extended lane blockage area LSA by connecting in the direction of lane travel.

In the above-mentioned surrounding object recognition action, as shown in FIG. 8, the own vehicle V traveling in the right lane LR changes lanes to the left lane LL in which two parked vehicles PV1 and PV2 form a parked lane. Assume a scene example similar to the background technology for generating the target trajectory Tt. In this case, the second parked vehicle PV2 in the parked convoy is a blind spot from the own vehicle V, but the first parked vehicle PV1 and the second parked vehicle PV2 in the parked convoy are combined into one. The parked vehicle line area PVA will be drawn based on the point group data (P1, ..., P9). Then, the lane blockage area LSA is set by expanding the drawn parked lane area PVA to the entire lane width of the left lane LL. Therefore, when generating the target trajectory Tt of the own vehicle V, the existence of the parked vehicle PV2, which is difficult to recognize the object due to hiding in the blind spot from the own vehicle V, is considered, and is ahead of the parked lane area PVA. Generates a target trajectory Tt that changes lanes to the left lane LL.

That is, in a scene where there are parked vehicles PV1 and PV2 that block the left lane LL of the own vehicle V traveling in the right lane LR, even if the parked vehicles PV1 and PV2 are not recognized separately, the right lane LR to the left lane The target trajectory Tt of the own vehicle V suitable for the lane change situation to LL will be generated. As a result, in various scenes where there are objects that block the planned driving lane of the own vehicle V, driving support suitable for the driving situation (traveling lane planning, target trajectory generation, etc.) without distinguishing and recognizing each object. Can be executed.

[Start point recognition action of lane blockage area (Figs. 9 and 10)]
The recognition action of the start point SP of the lane obstruction region LSA is a combination of the first start point recognition pattern based on the point cloud data Pn and the second start point recognition pattern based on the object recognition result. First, the first start point recognition pattern based on the point cloud data Pn will be described.

Of the first start point recognition pattern based on the point cloud data Pn acquired by the rider 113, in the case of the rear end of the parked vehicle, in the flowchart of FIG. 4, the lane blockage area is caused by the flow of S1 → S2 → S3 → S9. The starting point SP of LSA is recognized. Further, among the first start point recognition patterns based on the point cloud data Pn acquired by the rider 113, in the case of the construction section, in the flowchart of FIG. 4, the lane is due to the flow of S1 → S2 → S3 → S4 → S9. The starting point SP of the blocked area LSA is recognized.

In this first start point recognition pattern, as shown in FIG. 9, a plurality of bin division sections BD (n) are set by dividing the left lane LL in the lane traveling direction at regular intervals. Then, among the plurality of bin division sections BD (n), the bin division section BD (7) having a predetermined blockage width SW due to the obstacle OB in the lane width direction is generated by the aggregated point cloud data P1, P2, P2. , Recognized as a lane blockage candidate LSC.

Here, "binning" is one of the preprocessing often performed in machine learning and the like, and means to divide the data by a boundary and convert it into categorical data. The "blocking width SW" is SW = LW- (S1 + S2) when the lane width LW of the left lane LL, the width from the right boundary is the first space S1, and the width from the left boundary is the second space S2. Calculate with the formula. When the blockage width SW is equal to or greater than the obstacle determination threshold value, the bin division section BD (7) having the blockage width SW is recognized as a lane blockage candidate LSC. This is necessary for recognizing the start point SP of the lane blockage region LSA while reducing the calculation load compared to the process of identifying lane blockage candidates for all moving and stationary objects existing in the left lane LL. It will be possible to identify various lane blockage candidate LSCs.

Next, as shown in FIG. 10A, among the point cloud data Pn existing in the lane blockage candidate LSC, the data point at the position closest to the target locus Tt on which the own vehicle V is scheduled to travel is selected. , The point (point A) of the selected data point is recognized as the start point SP of the lane blockage region LSA. As a result, compared to the case where the starting point of the lane blockage region is set after grasping the shape of the rear end of the object by the point cloud data Pn, the number of scans for acquiring the point cloud data Pn by the rider 113 is reduced and the lane is set. The starting point SP of the closed area LSA can be set.

Here, as shown in FIG. 10B, the point cloud data Pn existing in the lane width direction from the start point SP of the lane blockage region LSA is connected to set the rear end surface region RA. Then, as shown in FIG. 10 (c), when the overlap amount OA between the set rear end surface region RA and the lane boundary on the side far from the own vehicle V is equal to or greater than a predetermined threshold value, the recognized start point SA is "parked". Set to "Starting point SA of lane blockage area LSA by vehicle PV". As a result, when recognizing the start point SA of the lane blockage area LSA based on the point cloud data Pn, the parked vehicle PV is separated from the obstacles existing on the road and "the start point SA of the lane blockage area LSA by the parked vehicle PV" is separated. "Can be set. That is, since the determination is made using the overlap amount OA of the rear end surface region RA, for example, an object other than the parked vehicle PV existing on the road, such as a signboard or a pole indicating the construction site, is regarded as the rear end portion of the parked vehicle PV. False recognition is suppressed.

Next, the second start point recognition pattern based on the object recognition result will be described. Among the second start point recognition patterns based on the object recognition result from the object recognition unit 315a, in the case of a parked vehicle, the flow of proceeding from S5 → S6 → S7 → S9 in the flowchart of FIG. 4 causes “lane blockage by the parked vehicle”. The starting point SP of the region LSA is recognized. Further, among the second start point recognition patterns based on the object recognition result from the object recognition unit 315a, in the case of the construction section, the flow proceeds from S5 → S6 → S7 → S8 → S9 in the flowchart of FIG. In the flowchart of 4, the "starting point SP of the lane blockage region LSA due to the construction section" is recognized by the flow of proceeding from S5 → S6 → S8 → S9.

As described above, in the second start point recognition pattern, if the object recognized by the object recognition process from the object recognition unit 315a is a stationary object that hinders the traveling of the own vehicle, whether or not the stationary object is a stationary vehicle. Is determined, and whether or not the stationary vehicle is a parked vehicle is determined. Then, when the vehicle is recognized as a parked vehicle, the position closest to the target trajectory of the own vehicle V in the rear end surface region of the vehicle is recognized as the "start point SP of the lane blockage region LSA by the parked vehicle". On the other hand, when it is determined to be a construction section, the position closest to the target trajectory of the own vehicle V in the rear end surface area of the object indicating the construction section is recognized as "the start point SP of the lane blockage region LSA due to the construction section". .. As a result, when recognizing the start point SP of the lane blockage region LSA, if the object recognition result can be obtained in advance, the start of the lane blockage region LSA using the object recognition result without using the point cloud data Pn from the rider 113. The point SP can be set.

In addition to the above contents, the control for recognizing the start point of the lane obstruction region is performed in consideration of the following points.
-When the vehicle is recognized as a parked vehicle, the starting point of the lane obstruction region may be recognized based on the front end surface region and the side surface region of the parked vehicle, not limited to the rear end surface region of the vehicle.
・ If there is a vehicle hazard, increase the probability of considering it as a parked vehicle.
・ If the right lane is crowded, do not increase the probability of considering it as a parked vehicle.
・ If the brake lamp of the vehicle is lit, reduce the probability of considering it as a parked vehicle.
・ Increase the probability of a parked vehicle when the amount of overlap on the left boundary of the vehicle is large. Furthermore, how to raise the probability is changed according to the road shoulder width.
-Increase the probability of recognizing a parked vehicle when it recognizes a triangular plate, smoke bomb, electronic signboard guide for police cars, traffic managers, etc. based on image processing.
-Recognize parked vehicles from the distribution of spatial location information on the map.
-Use communication to recognize parked vehicles.
・ At least a vehicle in a running state is not judged as a parked vehicle.
・ Increase the probability of a parked vehicle if the vehicle in front is stationary for a certain period of time and the right lane is open.
-If the lane blockade by the construction signboard or pylon is recognized based on the image processing, it will be the starting point of the lane blockade area.
-Recognize the start of the construction section from the distribution of spatial position information on the map.
・ If a surface that blocks the entire lane can be detected from the left road edge to the right side of the lane, it is recognized as a blockade due to the construction section.

[Lane blockage area expansion action by point cloud connection (Figs. 11 to 14)]
After the start point SP of the lane blockage area LSA is set, the expansion action of the lane blockage area LSA by connecting the point cloud data Pn from the rider 113 proceeds from S9 to S10 → S11 in the flowchart of FIG. → It is performed by repeating the flow of proceeding to S11.

First, in the expansion of the lane blockage region LSA, points that are points in the same lane and whose vertical deviation distance Lx that deviates from each other in the lane traveling direction is less than a predetermined threshold B are connected from the point cloud data Pn. Specifically, when the lane blockage area LSA is expanded starting from the start point SA of the lane blockage area LSA, as shown in FIG. 11, the point group data Pn to be connected are not adjacent and are separated from each other. The vertical deviation distances Lx1 and Lx2 at which the two data points P6 and P7 and the two data points P9 and P10 deviate in the lane travel direction are calculated. Then, for the vertical deviation distance Lx1, since the vertical deviation distance condition of less than the predetermined threshold B is satisfied, the two data points P6 and P7 by the first parked vehicle PV1 and the second parked vehicle PV2 are connected in the lane traveling direction. The lane blockage area LSA is expanded at. On the other hand, for the vertical deviation distance Lx2, since the vertical deviation distance condition exceeds the predetermined threshold B and the vertical deviation distance condition is not satisfied, the two data points P9 and P10 by the second parked vehicle PV2 and the temporarily stopped vehicle SV are not connected in the lane traveling direction. .. As a result, in the case of a parked convoy or a construction section in which a plurality of parked vehicles are lined up at a narrow inter-vehicle distance, the lane blockage area LSA can be expanded by connecting point clouds starting from the starting point SP. In addition, when the temporarily stopped vehicle SV exists at a position away from the leading vehicle in the parked convoy, the lane blockage area LSA up to the area that cannot be included in the lane blockage due to the failure of the vertical deviation distance condition. It will be possible to prevent the expansion of.

Next, when expanding the lane obstruction region LSA starting from the starting point SP, if the lateral deviation distance Ly between the point cloud data P3 to P6 and the point cloud data P7 to P10 is a predetermined value C or more, it is excluded from the connection target. .. Specifically, when the lane blockage region LSA is expanded starting from the start point SA of the lane blockage region LSA, as shown in FIG. 12, two point group data P3 to P6 and point group data P7 to P10 to be connected are to be connected. The maximum lateral deviation distance Ly, which is in the lane width direction and deviates toward the side approaching the driving lane (right lane LR) of the own vehicle V, is calculated. When the lateral deviation distance Ly is equal to or greater than the predetermined value C (margin in the lane width direction), the point group data P7 to P10 by the temporarily stopped vehicle SV are the parked vehicle PV regardless of whether the vertical deviation distance condition is satisfied or not. It is excluded from the connection target with the point group data P3 to P6 by. As a result, when the vertical divergence distance condition is satisfied, it is possible to prevent the temporarily stopped vehicle SV that may start moving from being mistaken as a parked vehicle PV and the lane blockage region LSA from being expanded by point cloud connection. That is, when there is a stationary vehicle that deviates from the side approaching the driving lane (right lane LR) of the own vehicle V, the stationary vehicle is not a parked vehicle PV but a temporary stop vehicle SV at the center position of the left lane LL. Probability is high.

Next, when expanding the lane blockage area LSA starting from the starting point SP, if the object attribute for which the point cloud data P10 and P11 scheduled to be connected next is acquired is a moving object, it is excluded from the connection target. Specifically, when the lane blockage area LSA is expanded starting from the start point SA of the lane blockage area LSA, as shown in FIG. 13, the point cloud data to be connected next to the point cloud data P1 to P9 is planned. The attributes of the object for which P10 and P11 have been acquired are determined. Then, when the attribute of the object is determined to be a stationary object (parked vehicle PV), the point cloud data P10 and P11 are linked. On the other hand, if the attribute of the object is determined to be a moving object (paused vehicle SV), the point cloud data P10 and P11 are excluded from the connection target. As a result, it is possible to prevent the temporarily stopped vehicle SV, which is likely to start moving, from being mistakenly recognized as the parked vehicle PV and expanding the lane blockage area LSA. As a result, it is possible to prevent the own vehicle V from accidentally changing lanes in front of the temporarily stopped vehicle SV and hindering the traveling of the temporarily stopped vehicle SV.

Next, when the lane blockage area LSA is expanded, when the starting point SP as the starting point cannot be recognized, the lane blockage area LSA (n) is expanded based on the past lane blockage area LSA (n-1). Specifically, as shown in FIG. 14, when the own vehicle V is traveling beside the parked convoy PV1, PV2, PV3, PV4 and the start point SP of the lane obstruction region LSA cannot be recognized, the lane obstruction occurs. When expanding the area LSA, if there is a previously recognized lane blockage area LSA (n-1), the past lane blockage area LSA (n-1) is used as the base. Then, the point group data P10 to P15 are connected in the lane traveling direction from the data point P9 of the past lane blockage region LSA (n-1) before the own vehicle V reaches the parked lane PV1, PV2, PV3, PV4. Expansion processing is performed to make the lane blockage area LSA (n) by one area. As a result, when the start point SP of the lane blockage area LSA cannot be recognized, the start point SP is replaced with the past lane blockage area LSA (n-1) and used to expand the lane blockage area LSA (n). Can be continued. As a result, in the case of a long parked vehicle line that exceeds the acquisition range of the point cloud data Pn by the rider 113 or in the case of a construction section over a long distance, the side of the already set lane blockage area LSA (n-1) When the own vehicle V travels, it is possible to set the lane blockage region LSA (n) by one region obtained by expanding the lane blockage region LSA (n-1).

In addition to the above contents, control is performed in consideration of the following points in the lane obstruction area expansion action by connecting point clouds.
・ When expanding the lane blockage area LSA, it is judged whether or not connection is possible based on the distribution of spatial position information.
-Spatial position information whose vertical distance in the lane is greater than or equal to the threshold value is not connected.
-Spatial position information (and surrounding spatial position information) whose position in the lateral direction of the lane is equal to or greater than the threshold value is not connected.
-If information on the type and state of the object can be obtained by the object recognition result, communication, or other means, it is associated with the spatial position information in the vicinity of the object and reflected in the judgment of connection availability.
-Dynamic change of the threshold value for determining whether or not to connect.
・ The threshold value for determining whether or not to connect is divided according to the cause of lane blockage. For example, if it is a construction section, it is assumed that the distance between pylon is long, so increase the distance threshold.
-When expanding the area by connecting the current spatial position information to the past lane blockage area, if a moving object is generated from the past lane blockage area, the past lane blockage area is not used.

As described above, the surrounding object recognition method and the surrounding object recognition device in the autonomous driving vehicle AD of the first embodiment have the effects listed below.

(1) A sensor (rider 113) that acquires point group data Pn (an example of spatial position information of an object) in the surrounding environment and a point group data Pn acquired by the sensor (rider 113) are used to surround the vehicle V. In a surrounding object recognition method including a controller (surrounding object recognition unit 315) that recognizes an existing object, the controller (surrounding object recognition unit 315) uses point group data Pn to cause an object (surrounding object recognition unit 315) that hinders the running of the own vehicle V. The lane blockage area LSA, which is the area where the lane (left lane LL) is blocked by the parked vehicle PV1 and PV2), is recognized, and the point group data Pn is connected in the lane traveling direction starting from the start point SP of the lane blockage area LSA. By doing so, the lane blockage area LSA is expanded (Fig. 8).
Therefore, in a scene where there are objects (parked vehicles PV1 and PV2) that block the planned lane of the own vehicle V, it is suitable for the driving situation even if the objects (parked vehicles PV1 and PV2) are not recognized separately. It is possible to provide a method of recognizing surrounding objects to perform driving support.

(2) When recognizing the starting point SP of the lane blockage area LSA, a plurality of bin division sections BD (n) are set by dividing the lane in the lane traveling direction at regular intervals, and a plurality of bin division sections BD (n) are set. Among them, the bin division section BD (7) having a predetermined blockage width SW due to the obstacle OB in the lane width direction is recognized as a lane blockage candidate LSC by the collected point group data Pn (FIG. 9).
Therefore, compared to the process of identifying lane blockage candidates for all moving objects and stationary objects existing in the planned traveling lane of the own vehicle V, the calculation load is reduced and the starting point SP of the lane blockage region LSA. It is possible to identify the lane blockage candidate LSC necessary for recognition of.

(3) From the point cloud data Pn existing in the lane blockage candidate LSC, select the data point P3 at the position closest to the target trajectory Tt of the own vehicle V, and set the point A of the selected data point P3 to the lane blockage area LSA. It is recognized as the starting point SP of (Fig. 10).
Therefore, compared to the case where the start point of the lane blockage region is set after grasping the shape of the rear end of the object from the point cloud data Pn, the number of scans for acquiring the point cloud data Pn is reduced and the lane blockage region LSA The starting point SP of can be set.

(4) The rear end face area RA is set by connecting the point group data Pn existing in the lane width direction from the start point SP of the lane blockage area LSA, and the set rear end face area RA and the lane boundary on the side far from the own vehicle V are set. When the amount of overlap OA with the above is equal to or greater than a predetermined threshold value, the recognized start point SP is set as the start point SP of the lane blockage region LSA by the parked vehicle PV (FIG. 10).
Therefore, when recognizing the start point SA of the lane blockage area LSA based on the point cloud data Pn, the parked vehicle PV is separated from the obstacles existing on the road and "the start point SA of the lane blockage area LSA by the parked vehicle PV" is separated. "Can be set.

(5) Exists on the traveling path of the own vehicle V based on the input information acquired from the in-vehicle sensor (camera 111, radar 112) including the spatial position information (point group data Pn) acquired by the sensor (rider 113). When the recognition process of the object is performed and the object recognized by the object recognition process is recognized as a stationary object (parked vehicle, construction section, etc.) that hinders the running of the own vehicle V, after the stationary object The position of the end face region closest to the target locus Tt of the own vehicle V is recognized as the start point SP of the lane obstruction region LSA (S5 to S9 in FIG. 4).
Therefore, when recognizing the start point SP of the lane blockage area LSA, if the object recognition result can be obtained in advance, the start point SP of the lane blockage area LSA is set using the object recognition result without using the point cloud data Pn. can do.

(6) When expanding the lane blockage area LSA starting from the starting point SP of the lane blockage area LSA, the vertical divergence distance Lx1 at which the two data points P6 and P7 to be connected deviate in the lane travel direction is calculated and the vertical divergence is calculated. When the vertical deviation distance condition that the distance Lx1 is less than the predetermined threshold B is satisfied, the two data points P6 and P7 are connected in the lane traveling direction to expand the lane blockage region LSA (FIG. 11).
Therefore, in the case of a parked convoy or a construction section, the lane blockage area LSA can be expanded by connecting the point cloud starting from the starting point SP, and the lane blockage area LSA to the area that cannot be included in the lane blockage can be expanded. Expansion can be prevented.

(7) When expanding the lane blockage area LSA starting from the start point SP of the lane blockage area LSA, the two point group data P3 to P6 and the point group data P7 to P10 to be connected are in the lane width direction and the own vehicle. Calculate the lateral divergence distance Ly that deviates toward the side approaching the driving lane of V, and if the lateral divergence distance Ly is equal to or greater than the predetermined value C, the point group data P7 to P10 are pointed regardless of whether the longitudinal divergence distance condition is satisfied or not. Exclude from the target of connection with group data P3 to P6 (Fig. 12).
Therefore, when the vertical deviation distance condition is satisfied, it is possible to prevent the temporarily stopped vehicle SV that may start moving from being mistaken for the parked vehicle PV and expanding the lane blockage region LSA by connecting the point cloud.

(8) When expanding the lane blockage area LSA starting from the start point SP of the lane blockage area LSA, the object attribute obtained from the data points P10 and P11 scheduled to be connected next is determined, and the object attribute is a stationary object. If it is determined that the object is a moving object, the point group data Pn is excluded from the connection target (FIG. 13).
Therefore, it is possible to prevent the temporarily stopped vehicle SV, which is likely to start moving, from being mistakenly recognized as the parked vehicle PV and expanding the lane blockage area LSA.

(9) When expanding the lane blockage area LSA in a scene where the start point SP of the lane blockage area LSA cannot be recognized, if the previously recognized lane blockage area LSA (n-1) exists, the past lane blockage area Based on LSA (n-1), the lane blockage region LSA (n) connecting the point group data Pn in the lane traveling direction is expanded (FIG. 14).
For this reason, in the case of a long parked vehicle line that exceeds the acquisition range of the point cloud data Pn or in the case of a construction section over a long distance, the own vehicle V is next to the already set lane blockage area LSA (n-1). When the vehicle travels, it is possible to set the lane obstruction region LSA (n) by one region which is an extension of the lane obstruction region LSA (n-1).

(10) A sensor (rider 113) that acquires point group data Pn (an example of spatial position information of an object) in the surrounding environment and a point group data Pn acquired by the sensor (rider 113) are used to surround the vehicle V. In a surrounding object recognition device including a controller (surrounding object recognition unit 315) for recognizing an existing object, the controller (surrounding object recognition unit 315) uses point group data Pn to cause an object (surrounding object recognition unit 315) that hinders driving of the own vehicle V. Lanes starting from the lane blockage area start point recognition unit 315b that recognizes the lane blockage area LSA, which is the area where the lane (left lane LL) is blocked by parked vehicles PV1 and PV2), and the start point SP of the lane blockage area LSA. It has a lane blockage region expansion portion 315c that expands the lane blockage region LSA by connecting the point group data Pn in the traveling direction (FIG. 3).
Therefore, in a scene where there are objects (parked vehicles PV1 and PV2) that block the planned lane of the own vehicle V, it is suitable for the driving situation even if the objects (parked vehicles PV1 and PV2) are not recognized separately. A surrounding object recognition device that executes driving support can be provided.

The surrounding object recognition method and the surrounding object recognition device of the present disclosure have been described above based on the first embodiment. However, the specific configuration is not limited to the first embodiment, and design changes and additions are permitted as long as the gist of the invention according to each claim of the claims is not deviated.

In Example 1, a rider 113 that acquires point cloud data Pn is used as a sensor that acquires spatial position information of an object in the surrounding environment. However, the sensor for acquiring the spatial position information of an object in the surrounding environment is not limited to the rider, and may be an example using a radar or a combination ranging sensor.

In Example 1, when the own vehicle V is traveling in the right lane LR and there is an obstacle in the left lane LL when changing lanes to the left lane LL, an example of setting the lane blockage area is shown. However, an example in which a lane blockage area is set when an obstacle exists in the driving lane of the own vehicle, and the own vehicle is changed lanes before and after the set lane blockage area so as to bypass the set lane blockage area. It may be. Furthermore, as an example of setting a lane block area when an obstacle exists in a lane adjacent to the driving lane of the own vehicle and determining a preceding vehicle when the own vehicle and another vehicle compete with each other in the section of the lane block area. Is also good.

In the first embodiment, an example is shown in which the surrounding object recognition method and the surrounding object recognition device of the present disclosure are applied to an automatically driving vehicle AD in which vehicle motion is controlled by speed and steering angle so as to travel along a target trajectory. .. However, the surrounding object recognition method and the surrounding object recognition device of the present disclosure are not limited to the autonomous driving vehicle, but are provided with an auto cruise function, a lane keeping function, and the like, and at least one of steering operation / accelerator operation / brake operation is performed. It can also be applied to driving support vehicles that support. In short, it can be applied to any vehicle equipped with a control system that uses information on surrounding objects in the lane in which the vehicle is traveling as input information in travel route generation, lane change control, and other controls.

AD Autonomous Driving Vehicle 11 External Sensor 111 Camera 112 Radar 113 Rider (Sensor)
31 Automatic operation controller 315 Surrounding object recognition unit (controller)
315a Object recognition unit 315b Lane blockage area start point recognition unit 315c Lane blockage area expansion unit 318 Travel area setting unit 319 Target trajectory generation unit

Claims (10)

A sensor that acquires spatial position information of objects in the surrounding environment,
In a surrounding object recognition method including a controller that recognizes an object existing around the own vehicle based on the spatial position information acquired by the sensor.
The controller
Based on the spatial position information, the lane obstruction area, which is an area where the lane is obstructed by an object that hinders the traveling of the own vehicle, is recognized.
A method for recognizing surrounding objects, which expands the lane obstruction region by connecting spatial position information of an object in the lane traveling direction starting from a start point of the lane obstruction region. In the surrounding object recognition method according to claim 1,
When recognizing the start point of the lane blockage region, a plurality of bin division sections are set by dividing the lane in the lane traveling direction at regular intervals.
Peripheral object recognition characterized in that, among the plurality of bin division sections, a bin division section having a predetermined blockage width due to an obstacle in the lane width direction is recognized as a lane blockage candidate based on the aggregated spatial position information. Method. In the surrounding object recognition method according to claim 2,
Among the spatial position information existing in the lane blockage candidate, the spatial position information at the position closest to the target trajectory of the own vehicle is selected.
A method of recognizing a surrounding object, which recognizes a point of the selected spatial position information as a starting point of the lane obstruction region. In the surrounding object recognition method according to claim 3,
The rear end surface region is set by connecting the spatial position information existing in the lane width direction from the start point of the lane blockage region.
When the amount of overlap between the set rear end surface region and the lane boundary on the side far from the own vehicle is equal to or greater than a predetermined threshold value, the recognized start point is set as the start point of the lane blockage region by the parked vehicle. Peripheral object recognition method. In the surrounding object recognition method according to any one of claims 1 to 4,
Based on the input information acquired from the in-vehicle sensor including the spatial position information acquired by the sensor, the recognition process of the object existing on the traveling path of the own vehicle is performed.
When the object recognized by the object recognition process is recognized as a stationary object that hinders the traveling of the own vehicle, the position closest to the target trajectory of the own vehicle in the rear end surface region of the stationary object is the lane. A method of recognizing surrounding objects, which is characterized by recognizing it as the starting point of a closed area. In the surrounding object recognition method according to any one of claims 1 to 5,
When the lane blockage area is expanded starting from the start point of the lane blockage area, the vertical dissociation distance at which the two spatial position information to be connected deviate in the lane traveling direction is calculated.
A method of recognizing a surrounding object, characterized in that when the vertical deviation distance condition that the vertical deviation distance is less than a predetermined threshold value is satisfied, the two spatial position information are connected in the lane traveling direction to expand the lane blockage region. In the surrounding object recognition method according to claim 6,
When the lane obstruction region is expanded starting from the start point of the lane obstruction region, the lateral divergence distance at which the two spatial position information to be connected are in the lane width direction and deviate toward the side approaching the traveling lane of the own vehicle is determined. Calculate and
A method of recognizing a surrounding object, characterized in that when the lateral deviation distance is equal to or greater than a predetermined value, the spatial position information is excluded from the connection target regardless of whether the vertical deviation distance condition is satisfied or not. In the surrounding object recognition method according to claim 6,
When expanding the lane blockage area starting from the start point of the lane blockage area, the object attribute for which the spatial position information scheduled to be connected is acquired is determined.
A method of recognizing a surrounding object, characterized in that when the object attribute is determined to be a stationary object, the spatial position information is connected, and when the object attribute is determined to be a moving object, the spatial position information is excluded from the connected object. .. In the surrounding object recognition method according to any one of claims 1 to 8.
When expanding the lane blockage area in a scene where the start point of the lane blockage area cannot be recognized, if the lane blockage area recognized in the past exists, the space is based on the past lane blockage area and is in the lane traveling direction. A method of recognizing surrounding objects, which comprises performing expansion processing of the lane blockage area for connecting position information. A sensor that acquires spatial position information of objects in the surrounding environment,
In a surrounding object recognition device including a controller that recognizes an object existing around the own vehicle based on the spatial position information acquired by the sensor.
The controller
The lane blockage area start point recognition unit that recognizes the lane blockage area, which is the area where the lane is blocked by an object that obstructs the traveling of the own vehicle, based on the spatial position information.
A peripheral object recognition device having a lane obstruction region expansion portion that expands the lane obstruction region by connecting spatial position information of an object in the lane traveling direction starting from a start point of the lane obstruction region. ..