Action plan generator

Abstract

To appropriately perform a traffic lane merge or the like.SOLUTION: An action plan generation device comprises: a storage unit 12 that stores a learning model generated by a model generation device; a detection unit 111 that detects traveling data for a self vehicle traveling on a first traffic lane; a recognition unit 112 that recognizes an action of another vehicle traveling on a second traffic lane which is adjacent to the first traffic lane or with which the first traffic lane merges; and an action plan generation unit 113 that generates an action plan for the self vehicle when the self vehicle moves from the first traffic lane to the second traffic lane on the basis of a level of yielding, which represents a possibility of the other vehicle giving way to the self vehicle indicated by an output value of the learning model when the traveling data for the self vehicle detected by the detection unit 111 and traveling data for the other vehicle based on the action of the other vehicle recognized by the recognition unit 112 are input to the learning model.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a behavior plan generating device that generates a behavior plan for a vehicle. [Background technology]

[0002] A known example of this type of device is one that, when performing driving control to cause a vehicle to change lanes or merge into lanes, determines whether or not the psychological burden on the vehicle's occupants will increase depending on the class of other vehicles traveling in the lane after the lane change or lane merging, and decides whether or not to execute driving control based on the results of this determination (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2021-28192 Summary of the Invention [Problem to be solved by the invention]

[0004] However, if the device described in Patent Document 1 determines whether or not to execute driving control based on the results of a judgment of the increase in the psychological burden on the occupants depending on the class of the other vehicle, the decision to execute driving control may be made before the driver of the other vehicle is fully aware of his or her own vehicle, and there is a risk that the driver will not be able to smoothly merge into other lanes, etc. [Means for solving the problem]

[0005] An action plan generation device according to one aspect of the present invention comprises: raw a detection unit that detects driving data of the vehicle traveling in the first lane; a recognition unit that recognizes the behavior of other vehicles traveling in the second lane adjacent to the first lane or merging with the first lane; and inputting the driving data of the vehicle detected by the detection unit and the driving data of other vehicles based on the behavior of the other vehicles recognized by the recognition unit into the learning model stored in the storage unit. The learning model predicts the distance between the vehicle and another vehicle at a predetermined time in the future based on the input driving data, estimates and outputs a yielding degree indicating the possibility that the other vehicle will yield to the vehicle based on the predicted distance, and calculates a yielding degree based on the outputted yielding degree.and a behavior plan generating unit that generates a behavior plan for the host vehicle when the host vehicle moves from the first lane to the second lane. Self vehicle and Other vehicles of The driving data includes at least one of position information and speed information. [Effects of the Invention]

[0006] According to the present invention, lane merging and the like can be performed smoothly. [Brief explanation of the drawings]

[0007] [Figure 1] 1 is a block diagram showing an outline of the overall configuration of a vehicle control system including a vehicle control device according to an embodiment of the present invention; [Figure 2] FIG. 2 is a block diagram illustrating an example of a main configuration of the vehicle control device of FIG. 1. [Figure 3A] FIG. 2 is a diagram showing an example of a driving scene of the host vehicle; [Figure 3B] FIG. 10 is a diagram showing another example of a driving scene of the host vehicle. [Figure 3C] FIG. 10 is a diagram showing another example of a driving scene of the host vehicle. [Figure 4] FIG. 2 is a block diagram illustrating an example of a configuration of a main part of the model generating device of FIG. 1. [Figure 5] FIG. 1 is a diagram for explaining generation of a learning model. [Figure 6A] 3 is a flowchart showing an example of processing executed by a controller of the model generating device of FIG. 2; [Figure 6B] 2 is a flowchart showing an example of processing executed by a controller of the vehicle control device of FIG. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, an embodiment of the present invention will be described with reference to FIGS. 1 to 6B. A vehicle control device having a behavior plan generation device according to an embodiment of the present invention can be applied to a vehicle having an automatic driving function, i.e., an automatic driving vehicle. A vehicle to which a vehicle control device according to the present embodiment is applied may be referred to as the host vehicle to distinguish it from other vehicles. The host vehicle may be an engine vehicle having an internal combustion engine (engine) as a driving source, an electric vehicle having a driving motor as a driving source, or a hybrid vehicle having an engine and a driving motor as driving sources. The host vehicle can run not only in an automatic driving mode in which no driving operation by the driver is required, but also in a manual driving mode in which the driver operates the vehicle.

[0009] 1 is a block diagram showing a schematic overall configuration of a vehicle control system 1 including a vehicle control device according to an embodiment of the present invention. The vehicle control system 1 includes a vehicle control device 100 and a model generation device 200 communicably connected to the vehicle control device 100 via a network NW. The network NW includes not only public wireless communication networks but also closed communication networks provided for each predetermined management area, such as wireless LAN, Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0010] Fig. 2 is a block diagram showing an example of a configuration of a main part of the vehicle control device 100 of Fig. 1. The vehicle control device 100 has a controller 10, a communication unit 13, a positioning unit 14, an internal sensor group 15, a camera 16, a navigation device 17, and a driving actuator AC. The vehicle control device 100 also has a behavior plan generation device 50 that constitutes a part of the vehicle control device 100. The behavior plan generation device 50 generates a driving trajectory (target trajectory) of the host vehicle from the present time to a predetermined time ahead, and generates a driving plan (behavior plan) according to the target trajectory.

[0011] The communication unit 13 communicates with the model generating device 200 via the network NW. The communication unit 13 also communicates with various servers via the network NW, and acquires map information, driving history information, traffic information, and the like from the servers periodically or at any timing. The acquired map information is output to the storage unit 12, and the map information is updated. The positioning unit (GNSS unit) 14 has a positioning sensor that receives positioning signals transmitted from positioning satellites. The positioning satellites are artificial satellites such as GPS satellites and quasi-zenith satellites. The positioning unit 14 measures the current position (latitude, longitude, altitude) of the vehicle using the positioning information received by the positioning sensor.

[0012] The internal sensor group 15 is a collective term for a plurality of sensors (internal sensors) that detect the driving state of the host vehicle. For example, the internal sensor group 15 includes a vehicle speed sensor that detects the vehicle speed of the host vehicle, an acceleration sensor that detects the longitudinal acceleration and the lateral acceleration (lateral acceleration) of the host vehicle, a rotation speed sensor that detects the rotation speed of the driving source, and a yaw rate sensor that detects the rotation angular velocity around the vertical axis of the center of gravity of the host vehicle. The internal sensor group 15 also includes sensors that detect the driver's driving operations in manual driving mode, such as operation of the accelerator pedal, operation of the brake pedal, and operation of the steering wheel. The camera 16 has an imaging element such as a CCD or a CMOS, and captures images of the surroundings (front, rear, and sides) of the host vehicle.

[0013] The navigation device 17 is a device that searches for a target route on roads to a destination input by the driver and provides guidance along the target route. The input of the destination and guidance along the target route are performed via an input / output device (not shown). The target route is calculated based on the current position of the vehicle measured by the positioning unit 14 and map information stored in the memory unit 12.

[0014] Actuators AC are driving actuators for controlling the driving of the host vehicle. When the driving source is an engine, actuators AC include a throttle actuator that adjusts the opening of the engine's throttle valve (throttle opening). When the driving source is a driving motor, actuators AC include the driving motor. Actuators AC also include a brake actuator that operates the host vehicle's braking device and a steering actuator that drives the steering device.

[0015] The controller 10 is composed of an electronic control unit (ECU). More specifically, the controller 10 includes a computer having an arithmetic unit 11 such as a CPU (microprocessor), a storage unit 12 such as a ROM and RAM, and other peripheral circuits (not shown) such as an I / O interface. Note that although multiple ECUs with different functions, such as an engine control ECU, a traction motor control ECU, and a braking device ECU, can be provided separately, for convenience, the controller 10 is shown in FIG. 2 as a collection of these ECUs.

[0016] The memory unit 12 stores the learning model generated by the model generating device 200. The memory unit 12 also stores highly accurate, detailed map information (referred to as high-accuracy map information). The high-accuracy map information includes road position information, road shape (curvature, etc.) information, road gradient information, intersection and branch point position information, number of lanes information, lane width and position information for each lane (information on lane center positions and lane boundary lines), position information of landmarks (traffic lights, signs, buildings, etc.) on the map, and road surface profile information such as road surface irregularities. The memory unit 12 also stores various control programs, thresholds used in the programs, and other information.

[0017] The calculation unit 11 has, as functional components, a detection unit 111, a recognition unit 112, a behavior plan generation unit 113, and a driving control unit 114. As shown in FIG. 1 , the detection unit 111, the recognition unit 112, and the behavior plan generation unit 113 are included in a behavior plan generation device 50.

[0018] The detection unit 111 detects driving data including at least the position information and speed information of the vehicle itself based on the current position of the vehicle measured by the positioning unit 14 and the sensor values ​​of the internal sensor group 15 .

[0019] The recognition unit 112 recognizes the external situation around the host vehicle. Specifically, the recognition unit 112 recognizes the behavior of vehicles around the host vehicle, more specifically, other vehicles traveling in lanes adjacent to the lane in which the host vehicle is traveling (hereinafter referred to as the host lane) (hereinafter referred to as the adjacent lane) or lanes into which the host lane merges (hereinafter referred to as the main lane), based on images captured by the camera 16. The recognition unit 112 also recognizes the road structure (road structure and shape) around the host vehicle based on the current position of the host vehicle measured by the positioning unit 14 and map information stored in the memory unit 12. The recognition unit 112 may also recognize the road structure around the host vehicle based on images captured by the camera 16. The recognition unit 112 may also recognize the road structure around the host vehicle based on information obtained through road-to-vehicle communication (V2I (Vehicle-to-Roadside-Infrastructure) communication) with a roadside device via the communication unit 13. Furthermore, the recognition unit 112 may recognize the behavior of other vehicles based on information obtained through vehicle-to-vehicle communication (V2V (Vehicle-to-Vehicle) communication) with other vehicles via the communication unit 13.

[0020] The behavior plan generation unit 113 generates a traveling trajectory (target trajectory) of the host vehicle from the present time to a predetermined time ahead based on, for example, a target route calculated by the navigation device 17, the current position of the host vehicle measured by the positioning unit 14, and the external conditions around the host vehicle recognized by the recognition unit 112. When there are multiple trajectories that are candidates for the target trajectory on the target route, the behavior plan generation unit 113 selects an optimal trajectory from among them that satisfies criteria such as compliance with laws and regulations and efficient and safe traveling, and sets the selected trajectory as the target trajectory. The behavior plan generation unit 113 then generates a behavior plan according to the generated target trajectory. The behavior plan generation unit 113 generates various behavior plans corresponding to traveling modes such as overtaking driving to overtake a preceding vehicle, lane-shifting driving to change the traveling lane to another lane (an adjacent lane or the main lane), following driving to follow a preceding vehicle, lane-keeping driving to maintain the traveling lane without deviating from the traveling lane, decelerating driving, or accelerating driving. When generating a target trajectory, the behavior plan generating unit 113 first determines a traveling mode, and generates the target trajectory based on the traveling mode.

[0021] In the automatic driving mode, the driving control unit 114 controls the actuators AC so that the vehicle travels along the target trajectory generated by the action plan generation unit 113. In the manual driving mode, the driving control unit 114 controls the actuators AC in response to a driving command (such as a steering operation) from the driver acquired by the internal sensor group 15.

[0022] In lane shifting driving, whether or not another vehicle traveling in a destination lane (an adjacent lane or the main lane) will yield to the host vehicle attempting to move into the other lane is determined by the intention of the driver of the other vehicle. Therefore, even if the driver of the other vehicle is fully aware of the host vehicle, the lane shifting driving may not be performed properly. For example, if the driver of the other vehicle decides not to yield to the host vehicle and performs lane shifting driving so that the host vehicle moves ahead of the other vehicle, sudden braking or sudden acceleration / deceleration control may be performed in the host vehicle and the other vehicle, which may cause anxiety or discomfort to the occupants of the host vehicle and the other vehicle. To address this problem, in this embodiment, the vehicle control device is configured as follows.

[0023] 3A, 3B, and 3C are diagrams showing examples of driving scenes of the host vehicle. FIG. 3A shows an example of a scene in which the host vehicle 101 is driving on a merging lane MG merging into a main lane ML with two lanes (lanes LN1 and LN2) on each side of left-hand traffic. FIG. 3B shows an example of a driving scene at a point in time after the point in time shown in FIG. 3A. In FIG. 3A, the host vehicle 101 is driving on the merging lane MG at a speed V1 per hour. Another vehicle 102 is driving on the left lane LN2 of the main lane ML at a speed V2 per hour, and another vehicle 103 is driving behind the other vehicle 102 at a speed V3 per hour. In the diagrams, D1 indicates the length of the tapered portion of the merging lane MG. D2 indicates the length of the section from the hard nose (the end of the guardrail GR) to the tip of the tapered portion of the merging lane MG (hereinafter referred to as the merging section). D3 indicates the length of the soft nose. D4 indicates the lane width of the left lane LN2 of the main road ML. D11 indicates the inter-vehicle distance (distance in the traveling direction) between the other vehicle 102 and the other vehicle 103.

[0024] In the driving scene shown in FIG. 3A, the detection unit 111 detects the current position of the host vehicle 101 measured by the positioning unit 14. The detection unit 111 also detects the driving speed (V1) of the host vehicle 101 based on the sensor value of the vehicle speed sensor of the internal sensor group 15. The recognition unit 112 recognizes that a merging path exists ahead of the host vehicle 101 in the traveling direction, where the host vehicle's lane (merging lane MG) and the main lane ML merge, based on the current position of the host vehicle measured by the positioning unit 14 and the map information stored in the memory unit 12. When the merging path is recognized, the recognition unit 112 detects the driving positions and driving speeds of surrounding vehicles based on the image captured by the camera 16, and recognizes surrounding vehicles whose inter-vehicle distance with the host vehicle 101 may be less than a predetermined distance when the host vehicle 101 moves onto the main lane ML, i.e., other vehicles involved in lane shift driving (other vehicles 102 and 103 in FIG. 3A). Furthermore, the recognition unit 112 recognizes the behavior of other vehicles involved in lane shifting based on the images captured by the camera 16.

[0025] The behavior plan generation unit 113 determines the traveling mode to be lane shift traveling when the recognition unit 112 recognizes a merging road ahead in the traveling direction of the host vehicle 101. As shown in FIG. 3B , when the positions in the traveling direction of the centers of gravity of all vehicles, the host vehicle 101 and the other vehicles 102, 103 involved in the lane shift traveling, exceed a hard nose, the behavior plan generation unit 113 estimates a yielding degree that indicates the possibility that the other vehicles 102, 103 will yield to the host vehicle 101.

[0026] More specifically, the behavior plan generation unit 113 inputs the travel data of the host vehicle 101 detected by the detection unit 111 and the travel data of the other vehicles 102, 103 based on the behavior of the other vehicles 102, 103 recognized by the recognition unit 112 into a learning model stored in the storage unit 12. The travel data of the other vehicles 102, 103 includes at least the position information and speed information of the other vehicles 102, 103. The learning model is configured using one or a combination of a rule base, a feed forward neural network (FFNN), and a recurrent neural network (RNN). Note that RNN includes LSTM (Long Short-Term Memory), which is an extension of RNN.

[0027] The learning model predicts the inter-vehicle distance between the host vehicle 101 and the other vehicles 102, 103 at a predetermined time in the future based on input travel data of the host vehicle 101 and travel data of the other vehicles 102, 103. The predetermined time in the future is the time when the host vehicle 101, which is a merging vehicle, reaches the end of the merging section, i.e., the tip of the tapered portion of the merging lane MG. The inter-vehicle distance at a predetermined time in the future can be calculated (predicted) using the formula Ix+Iv×t, where Ix is the inter-vehicle distance at the current time and Iv is the rate of change of the inter-vehicle distance. t is the elapsed time from the current time to a future time. The inter-vehicle distance at a predetermined time in the future may also be calculated (predicted) using the formula Ix+Iv×t+0.5(Ia×t^2). Ia is the rate of change of the rate of change of the inter-vehicle distance. 3C shows an example of a driving scene at a time point a predetermined time has elapsed since the time point in FIG. 3B, more specifically, at a predetermined time point in the future when the host vehicle 101 reaches the end of the merging section. D12 in the figure indicates the inter-vehicle distance between the host vehicle 101 and the other vehicle 103.

[0028] The learning model calculates (estimates) a yielding degree d, which indicates the likelihood that the other vehicles 102, 103 will yield to the host vehicle 101 when the host vehicle 101 approaches the other vehicles 102, 103, based on the calculated inter-vehicle distance, and outputs the calculation result. The yielding degree d is expressed as a numerical value between 0 (Low) and 1 (High). For example, d = 0.40 indicates that there is a 40% likelihood that the other vehicles 102, 103 will yield to the host vehicle 101. d = 0 indicates that there is no likelihood that the other vehicles 102, 103 will yield to the host vehicle 101, i.e., the other vehicles 102, 103 will not yield to the host vehicle 101. d = 1 indicates that the other vehicles 102, 103 will definitely yield to the host vehicle 101. Note that another vehicle (main line vehicle) that will be present behind the host vehicle 101 at a predetermined time in the future is the subject of calculation of the yielding degree d. Therefore, in the example shown in Fig. 3B, another vehicle 103 that will be present behind the host vehicle 101 at a predetermined time in the future is the subject of calculation of the yielding degree d.

[0029] The values ​​of D1 to D4 vary depending on the shape of the merging road. Furthermore, the timing at which the driver of another vehicle traveling on the main lane recognizes the subject vehicle and the timing at which the driver decides whether to yield to the subject vehicle also vary depending on the shape of the merging road. Therefore, the behavior plan generating unit 113 inputs the information of D1 to D4 into the learning model in addition to the driving data of the subject vehicle 101 and the driving data of the other vehicles 102 and 103 based on the behaviors of the other vehicles 102 and 103. As a result, the learning model outputs a yielding degree d corresponding to the road structure recognized based on D1 to D4. Note that, in addition to the above D1 to D4, information capable of identifying the road structure (such as the merging road) may also be input into the learning model.

[0030] The behavior plan generation unit 113 generates a target trajectory for the host vehicle 101 to move from the merging lane MG to the main lane ML based on the yielding degree d of the other vehicle (main lane vehicle) output by the learning model, and generates a behavior plan according to the target trajectory. More specifically, when the yielding degree d is equal to or greater than a predetermined threshold, the behavior plan generation unit 113 determines that the other vehicle will yield to the host vehicle 101, and generates a behavior plan such that the host vehicle 101 performs lane-changing driving, moving ahead of the other vehicle at a constant speed while maintaining its current cruising speed. On the other hand, when the yielding degree d is less than the predetermined threshold, the behavior plan generation unit 113 determines that the other vehicle will not yield to the host vehicle 101, and generates a behavior plan such that the host vehicle 101 performs lane-changing driving, moving behind the other vehicle while decelerating.

[0031] Next, the model generation device 200 will be described. FIG. 4 is a block diagram schematically showing an example of the configuration of the main parts of the model generation device 200 of FIG. 1. The model generation device 200 has a controller 20, a communication unit 23, an operation unit 24, a display 25, and a speaker 26. The model generation device 200 generates a learning model used for estimating the degree of yielding by the behavior plan generation unit 113 of the vehicle control device 100. The model generation device 200 has a simulator function that allows a simulated experience of driving a vehicle. Hereinafter, a user (driver) who experiences driving a vehicle using the model generation device 200 will be referred to as a subject.

[0032] The communication unit 23 communicates with external devices including the vehicle control device 100 via the network NW. The operation unit 24 has a seat on which the subject sits, and driving devices equivalent to those of an actual vehicle, such as an accelerator pedal, a brake pedal, and a steering wheel. When these driving devices are operated by the subject, driving commands such as acceleration / deceleration instructions and steering instructions are input to the controller 20. The display 25 is, for example, a liquid crystal display, and is installed in front of the seat so that the displayed image can be viewed by the subject seated in the seat. The display 25 displays an image (moving image) that simulates the view ahead as seen from the driver's seat of an actual vehicle, and the image changes according to the driving command input by the operation unit 24. The speaker 26 outputs sounds that simulate the driving sounds of an actual vehicle.

[0033] The controller 20 includes a computer having an arithmetic unit 21 such as a CPU (microprocessor), a storage unit 22 such as a ROM or RAM, and other peripheral circuits (not shown) such as an I / O interface.

[0034] The learning model is stored in the storage unit 22. The storage unit 22 also stores various control programs, threshold values ​​used in the programs, and other information.

[0035] The calculation unit 21 has, as its functional components, an environment setting unit 211, an output control unit 212, a data acquisition unit 213, and a model generation unit 214.

[0036] The environment setting unit 211 sets an experimental environment (simulation environment). Specifically, based on setting information input via the operation unit 24, it sets the structure and shape of the road (hereinafter referred to as a virtual road) on which the subject will experience driving, as well as a driving scenario. The setting information includes parameters for specifying the road structure (such as merging roads) and the road shape, such as D1 to D4 above. The driving scenario is information for determining the driving scene in which the subject will experience driving. By setting the driving scenario, the subject can experience driving a vehicle traveling on a main lane at a merging road (hereinafter referred to as a main lane vehicle), driving a vehicle moving from a merging lane to a main lane (hereinafter referred to as a merging vehicle), driving a vehicle moving from the current lane to an adjacent lane on a two-lane road, etc. The driving scenario includes information specifying the traffic conditions on the virtual road, more specifically, information specifying the number of other vehicles (hereinafter sometimes referred to as virtual vehicles) on the virtual road, and information specifying the behavior, driving position, driving speed, etc. of each virtual vehicle on the virtual road. In a driving scenario in which a user experiences driving a merging vehicle, the behavior of the virtual vehicle (main line vehicle) is specified, for example, as "accelerating and then decelerating due to being pulled by the merging vehicle," "continuing to accelerate due to being pulled by the merging vehicle," etc. In a driving scenario in which a user experiences driving a main line vehicle, the behavior of the virtual vehicle (merging vehicle) is specified, for example, as "entering the main line while traveling at a constant speed," "entering the main line while traveling at a reduced speed," etc.

[0037] The output control unit 212 generates moving image data for displaying moving images simulating the scenery seen from an actual vehicle on the display 25 based on driving commands input by the test subject operating the accelerator pedal, brake pedal, steering wheel, etc. of the operation unit 24, and outputs the generated moving image data to the display 25. In this way, driving and traveling equivalent to that of an actual vehicle are simulated. The output control unit 212 also generates audio data for outputting traveling sounds and the like corresponding to the simulated traveling, and outputs the generated audio data to the speaker 26.

[0038] The data acquisition unit 213 acquires driving data corresponding to the simulated driving and driving based on driving commands input by the subject via the operation unit 24. The data acquisition unit 213 also acquires driving data of the virtual vehicle based on the setting information input by the environment setting unit 211. The driving data acquired by the data acquisition unit 213 includes at least position information and speed information. When the subject's own vehicle (the vehicle in which the subject experiences driving) is a main line vehicle, the subject is instructed in advance whether or not to give way to a merging vehicle.

[0039] The model generation unit 214 generates a learning model based on the driving data acquired by the data acquisition unit 213. Here, the generation of the learning model by the model generation unit 214 will be described. Below, an example will be taken of a case where the structure of the virtual road is a merging road. Also, an example will be taken of a case where the learning model generated by the model generation unit 214 is rule-based. Furthermore, an example will be taken of a case where a test subject experiences driving a main line vehicle.

[0040] First, the data acquisition unit 213 acquires driving data of a main line vehicle whose driving is simulated by a test subject, and driving data of another vehicle (virtual vehicle) that traveled on a virtual road based on a driving scenario set by the environment setting unit 211. The model generation unit 214 generates a rule base that outputs a yielding degree that indicates the likelihood that a main line vehicle will yield to a merging vehicle, based on the driving data acquired by the data acquisition unit 213. The model generation unit 214 stores the generated rule base in the storage unit 22. The rule base is a learning model that outputs, as the yielding degree, a value obtained by multiplying a predetermined gain by the inter-vehicle distance between the main line vehicle and the merging vehicle at a predetermined future time point, assuming that the main line vehicle and the merging vehicle move at a constant velocity or a constant acceleration.

[0041] By repeatedly simulating the driving of main lane vehicles while changing the instructions given to the subjects regarding whether or not to give way to merging vehicles, the parameters of the rule base are updated and optimized.Furthermore, by repeatedly simulating the driving of main lane vehicles while changing the simulation environment, a rule base that corresponds to various traffic conditions and road structures is generated.

[0042] FIG. 5 is a diagram for explaining the generation of a learning model. Table DS shown in FIG. 5 is a table (data set) in which the driving data of the subject's own vehicle (the vehicle that the subject experienced driving) and other vehicles (virtual vehicles) that traveled on a virtual road are recorded in chronological order. Table DS is an example of a data set obtained when the subject experienced driving the same driving scene as in FIG. 3A on a virtual road having the same shape as the merging road in FIG. 3A, with the main lane vehicle (vehicle 102 in FIG. 3A) as the subject's own vehicle. Table DS includes driving data #1 of the subject's own vehicle (corresponding to vehicle 102 in FIG. 3A), driving data #2 of a virtual vehicle (corresponding to vehicle 101 in FIG. 3A), and driving data #3 of a virtual vehicle (corresponding to vehicle 103 in FIG. 3A). Driving data #1 includes position information X11(t) indicating the position of the subject's own vehicle at time t and speed information V11(t) indicating the driving speed of the subject's own vehicle at time t. The running data #2 and #3 include position information X12(t) and X13(t) indicating the position of each virtual vehicle at time t, and speed information V12(t) and V13(t) indicating the running speed of each virtual vehicle at time t.

[0043] When the driving data #1 to #3 are input to the learning model LM, the degree of yielding d is calculated (estimated) by the learning model LM. The characteristic f1 in FIG. 5 represents the output value of the learning model LM at each time, that is, the degree of yielding d. Note that at time t m is the time when the positions of the centers of gravity of all vehicles, including the host vehicle and the virtual vehicle, in the direction of travel pass the hard nose on the virtual road, and is the time when estimation of the yielding degree d by the learning model LM begins.

[0044] Fig. 6A is a flowchart showing an example of processing executed in accordance with a predetermined program by the controller 10 of the model generation device 200 of Fig. 2. The processing shown in the flowchart of Fig. 6A is started, for example, when a command to start a simulation is input by the subject via the operation unit 24.

[0045] First, in step S11, a simulation environment is set based on the setting information input via the operation unit 24. In step S12, a simulation is started based on the simulation environment set in step S11. Specifically, a moving image corresponding to the simulation environment set in step S11 (a moving image simulating the scenery seen from an actual vehicle) is displayed on the display 25, and acceptance of driving commands from the test subject is started. When a driving command from the test subject is input via the operation unit 24, moving image data and audio data corresponding to the driving command are generated. The generated moving image data is then output to the display 25, and the generated audio data is output to the speaker 26. This simulates driving and traveling equivalent to that of an actual vehicle.

[0046] In step S13, it is determined whether the merging vehicle has completed lane change. Whether lane change is complete is determined based on whether the merging vehicle that entered the main lane has reached the end of the merging section. Step S13 is repeated until the result is affirmative.

[0047] If the result in step S13 is affirmative, then in step S14, travel data (for example, travel data #1 to #3 shown in FIG. 5) is acquired. In step S15, a learning model is generated based on the travel data acquired in step S14. The generated learning model is stored in the storage unit 22.

[0048] 6B is a flowchart showing an example of processing executed by the controller 10 of the vehicle control device 100 of FIG. 1 in accordance with a predetermined program. The processing shown in the flowchart of FIG. 6B is repeated at predetermined intervals while the host vehicle 101 is traveling in autonomous driving mode, for example. Here, it is assumed that the host vehicle 101 is traveling in autonomous driving mode in the merging lane MG of FIG. 3A. It is also assumed that the learning model generated by the model generation device 200 is acquired by the vehicle control device 100 via the network NW and is stored in advance in the storage unit 12 of the vehicle control device 100. Note that a user may transfer the learning model from the model generation device 200 to the vehicle control device 100 using a recording medium or the like.

[0049] First, in step S21, the road structure ahead in the traveling direction of the host vehicle 101 is recognized based on the current position of the host vehicle measured by the positioning unit 14 and the map information stored in the memory unit 12. Furthermore, parameters (D1 to D4 above) representing the shape of the road are acquired based on the recognition result. In step S22, travel data of the host vehicle 101 is detected based on the current position of the host vehicle measured by the positioning unit 14 and the sensor values ​​of the internal sensor group 15. In step S23, the behavior of the other vehicles 102, 103 is recognized based on the image captured by the camera 16. In step S24, the travel data of the host vehicle 101 detected in step S22 and travel data based on the behavior of the other vehicles recognized in step S23 are input to a learning model stored in the memory unit 12. At this time, the information representing the road structure recognized in step S21 and the parameters representing the shape of the road acquired in step S21 are also input to the learning model.

[0050] In step S25, it is determined whether the output value of the learning model, i.e., the yielding degree d, is equal to or greater than a predetermined threshold, i.e., whether the other vehicle 103 will yield to the host vehicle 101. If the result in step S25 is affirmative, in step S26, an action plan is generated so that the host vehicle 101 moves between lanes at a constant speed to move ahead of the other vehicle 103. At this time, the action plan may be generated so that the host vehicle 101 accelerates at a greater acceleration as the yielding degree d approaches the predetermined threshold. As a result, even if the other vehicle 103 accelerates just before the host vehicle 101 moves onto the main lane, the host vehicle 101 can move ahead of the other vehicle 103 while maintaining a sufficient inter-vehicle distance, and sudden braking or sudden acceleration to avoid a collision with the other vehicle 103 can be suppressed.

[0051] If the result in step S25 is negative, an action plan is generated in step S27 so that the host vehicle 101 performs lane change traveling while decelerating to move behind the other vehicle 103. At this time, the action plan may be generated so that the host vehicle 101 decelerates at a greater deceleration as the yielding degree d approaches a predetermined threshold. As a result, even if the other vehicle 103 decelerates just before the host vehicle 101 moves onto the main lane, the host vehicle 101 can move ahead of the other vehicle 103 while maintaining a sufficient inter-vehicle distance, and sudden braking or sudden deceleration to avoid a collision with the other vehicle 103 can be suppressed.

[0052] According to this embodiment, the following effects can be achieved. (1) The device includes a memory unit 12 that stores a learning model generated by the model generation device 200, a detection unit 111 that detects driving data of the host vehicle 101 driving in the first lane (merging lane MG in Figure 3A), a recognition unit 112 that recognizes the behavior of other vehicles 102, 103 driving in the second lane (lane LN2 in Figure 3A) adjacent to the first lane or into which the first lane merges, and a behavior plan generation unit 113 that generates a behavior plan for the host vehicle 101 when the host vehicle 101 moves from the first lane to the second lane based on the output value of the learning model when the driving data of the host vehicle 101 detected by the detection unit 111 and the driving data of the other vehicles based on the behavior of the other vehicles 102, 103 recognized by the recognition unit 112 are input into the learning model stored in the memory unit 12. The model generation device 200 includes a data acquisition unit 213 that acquires driving data including at least one of position information and speed information of a first vehicle when a first vehicle traveling in a first lane moves to a second lane, and driving data including at least one of position information and speed information of a second vehicle traveling in the second lane, based on the driving data acquired by the data acquisition unit 213, and a model generation unit 214 that generates a learning model that calculates and outputs a yielding degree that indicates the likelihood that the second vehicle will yield to the first vehicle when the first vehicle approaches the second vehicle. The driving data of the host vehicle 101, the other vehicles 102 and 103, the first vehicle, and the second vehicle includes at least one of position information and speed information. The learning model includes one or a combination of a rule-based, FFNN, and RNN. This generates a behavior plan that allows the host vehicle to smoothly move between lanes in accordance with the behavior of the other vehicles, thereby enabling smooth lane merging and other operations.

[0053] (2) When the learning model is rule-based, the behavior plan generation unit 113 inputs the traveling data acquired by the data acquisition unit 213 into the rule base. The rule base predicts the inter-vehicle distance between the host vehicle 101 and the other vehicles 102, 103 at a predetermined time in the future based on the traveling data input by the behavior plan generation unit 113, and estimates and outputs a yielding degree indicating the likelihood that the other vehicles 102, 103 will yield to the host vehicle 101 based on the predicted inter-vehicle distance. This allows the host vehicle 101 to change lanes taking into account the inter-vehicle distance from the other vehicles 102, 103, thereby enabling safer lane-changing driving.

[0054] (3) The recognition unit 112 further recognizes the road structure around the host vehicle 101. Based on the recognition result of the road structure by the recognition unit 112, the behavior plan generation unit 113 acquires parameters (D1 to D4 in FIG. 3A) representing the road structure and inputs them into the rule base. When the first lane in which the host vehicle 101 is traveling is a merging lane (merging lane MG in FIG. 3A) that merges into the second lane, the rule base calculates the remaining distance until the host vehicle 101 reaches the end of the tapered portion of the merging lane based on the parameters input by the behavior plan generation unit 113, and predicts the inter-vehicle distance between the host vehicle 101 and other vehicles 102, 103 at a predetermined future time when the host vehicle 101 will reach the end of the tapered portion of the merging lane based on the calculated remaining distance. This allows the inter-vehicle distance between the host vehicle 101 and other vehicles 102, 103 when the host vehicle 101 reaches the end of the tapered portion of the merging lane to be accurately calculated. As a result, lane merging and other operations can be performed smoothly and safely regardless of the structure or shape of the road on which the vehicle is traveling.

[0055] (4) The behavior plan generation unit 113 generates a behavior plan so that the acceleration / deceleration of the vehicle when entering lane 2 from lane 1 increases or decreases according to the output value of the learning model. This suppresses sudden braking and sudden acceleration / deceleration to avoid collisions with other vehicles when moving between lanes, thereby improving traffic safety.

[0056] (5) The data acquisition unit 213 acquires driving data of the second vehicle, which is specified in advance as the second vehicle's driving data, which is either a first behavior in which the second vehicle gives way to the first vehicle, or a second behavior in which the second vehicle does not give way to the first vehicle.

[0057] The above embodiment can be modified in various ways. Some modified examples will be described below. In the above embodiment, the data acquisition unit 213 as an acquisition unit acquires, as the driving data of the second vehicle, driving data in which it is specified in advance whether the second vehicle will take a first behavior of yielding to the first vehicle or a second behavior of not yielding to the first vehicle. However, the acquisition unit may acquire, as the driving data of the second vehicle, driving data in which it is specified in advance whether the second vehicle will take a first behavior of yielding to the first vehicle, a second behavior of not yielding to the first vehicle, or a third behavior of leaving it up to the driver to decide whether to yield to the first vehicle.

[0058] In the above embodiment, when the first lane on which the host vehicle 101 is traveling is a merging lane that merges into the second lane, the rule base calculates the remaining distance until the host vehicle 101 reaches the end of the tapered portion of the merging lane, and predicts the inter-vehicle distance between the host vehicle 101 and other vehicles 102, 103 at a predetermined future time (the time when the host vehicle 101 reaches the end of the tapered portion of the merging lane) based on the calculated remaining distance. However, when the first lane on which the host vehicle 101 is traveling is an adjacent lane adjacent to the second lane, the host vehicle 101 may perform lane change traveling from the first lane to the second lane. Therefore, when the first lane on which the host vehicle 101 is traveling is an adjacent lane adjacent to the second lane, the rule base may calculate the distance in the traveling direction at which the host vehicle 101 can change lanes from the first lane to the second lane (hereinafter referred to as the lane change distance) based on parameters that specify the structure and shape of the virtual road, which are included in the setting information. The lane change distance is, for example, the distance until the host vehicle reaches another vehicle or obstacle that must be avoided ahead in the first lane in which the host vehicle is traveling. When the object to be avoided is another vehicle, the lane change distance may be calculated taking into account the traveling speed of the other vehicle. The rule base may predict the inter-vehicle distance between the host vehicle 101 and another vehicle traveling in the second lane at a predetermined future time point (the time point when the host vehicle 101 reaches the object to be avoided) based on the calculated lane change distance.

[0059] Furthermore, in the above embodiment, the detection unit 111 detects driving data including at least the position information and speed information of the host vehicle. However, the detection unit may detect driving data including at least one of the position information, speed information, and acceleration information of the host vehicle. Furthermore, in the above embodiment, the behavior plan generation unit 113 inputs driving data of other vehicles including at least the position information and speed information of the other vehicles to the learning model. However, the behavior plan generation unit may input driving data of other vehicles including at least the position information, speed information, and acceleration information of the other vehicles to the learning model. Furthermore, in the above embodiment, the data acquisition unit 213 as an acquisition unit acquires driving data including at least the position information and speed information. However, the acquisition unit may acquire driving data including at least the position information, speed information, and acceleration information.

[0060] In the above embodiment, the model generation unit 214 generates a rule base using driving data obtained through experiments by subjects as training data. However, the same applies to generating other learning models. For example, the model generation unit may use driving data obtained through experiments by subjects as training data to generate an FFNN or LSTM that recognizes patterns of behavior of other vehicles. In this case, the behavior plan generation unit inputs the driving data of the host vehicle detected by the detection unit and the driving data of other vehicles based on the behavior of the other vehicles recognized by the recognition unit into the FFNN or LSTM, and estimates the degree to which other vehicles will yield to the host vehicle based on the behavior pattern of the other vehicles recognized by the FFNN or LSTM.

[0061] In the above embodiment, the model generation device 200 is connected to the vehicle control device 100 via a network NW. However, the model generation unit of the model generation device may be included in the vehicle control device. More specifically, the calculation unit of the vehicle control device may have the model generation unit as a functional component. In this case, the model generation unit of the vehicle control device communicates with the model generation device via a communication unit to acquire driving data acquired by a data acquisition unit of the model generation device, and generates a learning model using the driving data as training data. Note that a user may transfer the driving data acquired by the data acquisition unit of the model generation device from the model generation device to the vehicle control device using a recording medium or the like. In addition, the model generation unit of the vehicle control device may generate a learning model using training data of the vehicle itself detected by the detection unit and driving data of other vehicles based on the behavior of the other vehicles recognized by the recognition unit.

[0062] In the above embodiment, the behavior plan generation unit 113 generates a behavior plan for the host vehicle 101 so that the host vehicle 101 moves ahead of the other vehicle when the yielding degree d is equal to or greater than a predetermined threshold. However, for example, in the traveling scene of Fig. 3C, when it is predicted that the inter-vehicle distance between the other vehicles 102 and 103 will be less than a predetermined length, the behavior plan generation unit may generate a behavior plan for the host vehicle 101 so that the host vehicle 101 moves behind the other vehicle 103. This allows the host vehicle 101 to move in a lane more safely.

[0063] Furthermore, in the above embodiment, the vehicle control device 100 is applied to an autonomously driven vehicle, but the vehicle control device 100 can also be applied to vehicles other than autonomously driven vehicles. For example, the vehicle control device 100 can also be applied to a manually driven vehicle equipped with an ADAS (Advanced Driver-Assistance Systems).

[0064] The above description is merely an example, and the present invention is not limited to the above-described embodiment and modifications, as long as the features of the present invention are not impaired. One or more of the above-described embodiment and modifications can be arbitrarily combined, and modifications can also be combined with each other. [Explanation of symbols]

[0065] 10 Controller, 12 Memory unit, 13 Communication unit, 14 Positioning unit, 15 Internal sensor group, 16 Camera, 17 Navigation device, 111 Detection unit, 112 Recognition unit, 113 Action plan generation unit, 114 Travel control unit, 20 Controller, 22 Memory unit, 23 Communication unit, 24 Operation unit, 25 Display, 26 Speaker, 211 Environment setting unit, 212 Output control unit, 213 Data acquisition unit, 214 Model generation unit, AC actuator

Claims

1. A memory unit that stores the generated learning model; a detection unit that detects travel data of a host vehicle traveling in a first lane; a recognition unit that recognizes the behavior of another vehicle traveling in a second lane adjacent to the first lane or merging with the first lane; and a behavior plan generation unit that inputs travel data of the host vehicle detected by the detection unit and travel data of the other vehicle based on the behavior of the other vehicle recognized by the recognition unit into the learning model stored in the storage unit, the learning model predicting an inter-vehicle distance between the host vehicle and the other vehicle at a predetermined time in the future based on the input travel data, estimating and outputting a yielding degree indicating a possibility that the other vehicle will yield to the host vehicle based on the predicted inter-vehicle distance, and generating a behavior plan for the host vehicle when moving from the first lane to the second lane based on the outputted yielding degree, The traveling data of the host vehicle and the other vehicle includes at least one of position information and speed information. Action plan generator.

2. 2. The behavior plan generating device according to claim 1, The learning model stored in the storage unit includes: The neural network is characterized by including any one or a combination of a rule-based, a feedforward neural network, and a recurrent neural network. Action plan generator.

3. In the behavior plan generation device according to claim 1, The learning model is rule-based. Action plan generator.

4. 4. The behavior plan generating device according to claim 3, The recognition unit further recognizes a road structure around the vehicle, the action plan generation unit acquires parameters representing the road structure based on the recognition result of the road structure by the recognition unit, and inputs the parameters into the rule base; The rule base is characterized in that, when the first lane in which the host vehicle is traveling is a merging lane merging into the second lane, the rule base calculates a remaining distance until the host vehicle reaches an end of a tapered portion of the merging lane based on the parameters input by the action plan generation unit, and predicts the inter-vehicle distance between the host vehicle and the other vehicle at the predetermined future time when the host vehicle will reach the end of the tapered portion of the merging lane based on the calculated remaining distance. Action plan generator.

5. 5. The behavior plan generating device according to claim 1, the behavior plan generation unit generates the behavior plan so that acceleration / deceleration of the host vehicle when moving from the first lane to the second lane increases or decreases according to an output value of the learning model. Action plan generator.

6. A memory unit that stores a learning model generated by the model generation device; a detection unit that detects travel data of a host vehicle traveling in a first lane; a recognition unit that recognizes the behavior of another vehicle traveling in a second lane adjacent to the first lane or merging with the first lane; a behavior plan generation unit that generates a behavior plan for the host vehicle when the host vehicle moves from the first lane to the second lane, based on an output value of the learning model stored in the storage unit when the running data of the host vehicle detected by the detection unit and the running data of the other vehicle based on the behavior of the other vehicle recognized by the recognition unit are input to the learning model, The model generating device an acquisition unit that acquires travel data of the first vehicle traveling on the first lane and travel data of a second vehicle traveling on the second lane when the first vehicle traveling on the first lane moves to the second lane; a generation unit that generates the learning model by calculating a yielding degree that indicates a possibility that the second vehicle will yield to the first vehicle when the first vehicle approaches the second vehicle, based on the traveling data acquired by the acquisition unit, and outputs the calculated yielding degree; the acquisition unit acquires, as the traveling data of the second vehicle, traveling data of the second vehicle in which a first behavior of yielding to the first vehicle and a second behavior of not yielding to the first vehicle are specified in advance; The traveling data of the host vehicle, the other vehicle, the first vehicle, and the second vehicle includes at least one of position information and speed information. Action plan generator.

7. A memory unit that stores a learning model generated by the model generation device; a detection unit that detects travel data of a host vehicle traveling in a first lane; a recognition unit that recognizes the behavior of another vehicle traveling in a second lane adjacent to the first lane or merging with the first lane; a behavior plan generation unit that generates a behavior plan for the host vehicle when the host vehicle moves from the first lane to the second lane, based on an output value of the learning model stored in the storage unit when the running data of the host vehicle detected by the detection unit and the running data of the other vehicle based on the behavior of the other vehicle recognized by the recognition unit are input to the learning model, The model generating device an acquisition unit that acquires travel data of the first vehicle traveling on the first lane and travel data of a second vehicle traveling on the second lane when the first vehicle traveling on the first lane moves to the second lane; a generation unit that generates the learning model by calculating a yielding degree that indicates a possibility that the second vehicle will yield to the first vehicle when the first vehicle approaches the second vehicle, based on the traveling data acquired by the acquisition unit, and outputs the calculated yielding degree; the acquisition unit acquires, as the traveling data of the second vehicle, traveling data of the second vehicle in which any one of a first behavior of yielding to the first vehicle, a second behavior of not yielding to the first vehicle, and a third behavior of leaving it up to the driver to decide whether to yield to the first vehicle is to be assumed; The traveling data of the host vehicle, the other vehicle, the first vehicle, and the second vehicle includes at least one of position information and speed information. Action plan generator.

Images