Driving scenario search device and simulation system

Abstract

To search for a traveling scenario that is desired to be verified in advance from a traffic flow that is simulated by simulation.SOLUTION: A traveling scenario search device includes: a traffic flow simulator 30 that simulatively generates a traffic flow on a road network; a traffic flow log data generation unit 222 that generates time-series data representing behavior of each vehicle model included in the traffic flow generated by the traffic flow simulator 30 as traffic flow log data; and a traffic flow log data analysis unit 223 that analyzes the traffic flow log data generated by the traffic flow log data generation unit 222 and searches for a specific vehicle model that matches a specific traveling scenario from vehicle models included in the traffic flow.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a driving scenario search device and a simulation system, and more particularly to a search device that finds a vehicle model that matches a specific driving scenario from traffic flows generated by simulation, and a simulation system that replaces the searched vehicle model with a vehicle model that simulates an automatic driving function and runs the vehicle in a simulated manner. [Background technology]

[0002] Automated lane keeping systems (ALKS) are known to keep a vehicle within its lane and travel at a specified speed or less by controlling the vehicle's forward, backward, and lateral movements over a long period of time. ALKS is realized by combining vehicle position recognition, vehicle speed and route target generation, vehicle speed control, route following control, and automatic steering. Verification of the functions and performance of advanced driver assistance systems (ADAS) and automated driving (AD) systems, including ALKS, can be achieved through test course and on-road testing, as well as through simulation tools that perform driving simulations of vehicle models equipped with ADAS or AD control models. For example, in driving simulations, a traffic flow model that models the behavior of multiple traffic participants is generated, and a vehicle model equipped with an ADAS or AD control model is driven through the generated traffic flow.

[0003] For example, Patent Document 1 discloses a driving simulation system that uses simulations to confirm the functionality of autonomous driving in various situations. The system manages the state transitions from the initial state of each other vehicle based on a driving scenario that includes the initial states of multiple other vehicles relative to a reference vehicle and the operation definitions of each other vehicle, and controls traffic flow within the simulation using a large number of other vehicles. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-117329 Summary of the Invention [Problem to be solved by the invention]

[0005] Possible methods for verifying the functions and performance of ADAS or AD through simulation include creating a scenario in which the vehicle speeds, positions, routes, etc. of other traffic participants are specified in advance, as in Patent Document 1, and running a vehicle model equipped with an ADAS or AD control model within the specified scenario; and generating a random traffic flow by defining only the number of other traffic participants appearing per hour, their speed ranges, etc., and running a vehicle model equipped with an ADAS or AD control model within that flow.

[0006] To verify the functionality of ADAS or AD in specific driving scenarios, such as following a vehicle ahead or overtaking a vehicle, the driving scenario to be verified must be included in the traffic flow simulated by the simulation tool. However, as described above, in the method of generating random traffic flow, the behavior of each traffic participant changes depending on their mutual relationships, making it impossible to predict in advance what kind of driving scenario will be generated. Therefore, even if a simulation is performed, there may be cases where the driving scenario to be verified is not included in the generated traffic flow.

[0007] The present invention has been made in view of the above-described circumstances, and its purpose is to search for a driving scenario that is to be verified in advance from a traffic flow that is generated artificially by a simulation. [Means for solving the problem]

[0008] According to one aspect of the present invention, a driving scenario search device includes a traffic flow simulator that simulates traffic flow on a road network, a traffic flow log data generation unit that generates time-series data representing the behavior of each vehicle model included in the traffic flow generated by the traffic flow simulator as traffic flow log data, and a traffic flow log data analysis unit that analyzes the traffic flow log data generated by the traffic flow log data generation unit and searches for a specific vehicle model that matches a specific driving scenario from among the vehicle models included in the traffic flow. According to one aspect of the present invention, a simulation system for theoretically verifying a vehicle's autonomous driving function comprises the above-mentioned driving scenario search device, a verification simulator that generates a verification vehicle model that simulates the autonomous driving function, and a management device that operates the verification simulator and the traffic flow simulator in conjunction with each other, wherein the management device designates the specific vehicle model searched for by the driving scenario search device as a replacement vehicle model, and replaces the verification vehicle model of the verification simulator with the replacement vehicle model in the traffic flow generated by the traffic flow simulator and simulates driving the vehicle. [Effects of the Invention]

[0009] According to the present invention, it is possible to search for a driving scenario that one wishes to verify in advance from a simulated traffic flow. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing a schematic configuration of a simulation system according to an embodiment of the present invention. [Figure 2] 2(a) to 2(d) are diagrams showing examples of driving scenarios. [Figure 3] FIG. 3 is a flowchart illustrating the flow of processing in the simulation system. [Figure 4] 4(a) to 4(d) are diagrams illustrating examples of reference scenario patterns for determining a specific driving scenario. [Figure 5]5(a) to 5(d) are diagrams illustrating a method for searching for a vehicle model that matches a specific driving scenario. DETAILED DESCRIPTION OF THE INVENTION

[0011] In one embodiment of the present invention described below, a method for verifying the function and performance of an advanced driver assistance system (ADAS) or autonomous driving (AD) that has an automated lane keeping system (ALKS) that automatically controls the forward / backward (longitudinal) and left / right (lateral) movements of a vehicle to keep it in lane, for example, is used in which a vehicle model equipped with an ADAS or AD control model is driven in a random traffic flow by simulation. Here, the advanced driver assistance system (ADAS) can be considered as part of the concept / function of autonomous driving (AD), so hereinafter both are collectively referred to as the autonomous driving function.

[0012] Specifically, a traffic flow model is set up in advance with a range of parameters for the behavior of other traffic participants, such as the number of vehicles appearing per hour, vehicle speed, and following distance when following a vehicle in front, to simulate the generation of random traffic flow. Then, a vehicle model equipped with a control model for autonomous driving functions is simulated to drive through the generated random traffic flow, thereby conducting a desktop verification of the functionality and performance of the autonomous driving control system. Here, random traffic flow refers to traffic flow in which each vehicle behaves randomly within the set parameter ranges, and it is not possible to predict in advance how each vehicle will behave. In random traffic flow, each vehicle model behaves freely within the set parameter ranges.

[0013] Random traffic flows can be generated, for example, by modeling traffic participants as particles and calculating microvariables, including the position and speed of each traffic participant, based on the interrelationships between them. Traffic flows generated in this way are called microtraffic flows. Individual vehicles in the microtraffic flow move randomly according to the calculated microvariables, influenced by the surrounding environment, such as the road network, traffic signals, and the behavior of other traffic participants.

[0014] In other words, in random traffic flow, each vehicle model does not behave in a unique way that is predetermined, but generates traffic flow by driving while maintaining interrelationships with other vehicle models within the ranges of set parameters such as vehicle speed, forward and rearward visibility distance, inter-vehicle distance, deceleration, etc. Therefore, even if the set parameter ranges are the same, each vehicle model behaves freely within the parameter ranges, resulting in different traffic flows being generated for each simulation.

[0015] For example, to verify the behavior of a vehicle model equipped with a control model for autonomous driving functions in a specific driving scenario, such as overtaking, the driving scenario to be verified must be included in the traffic flow generated by the simulator.However, when a vehicle model is run in a random traffic flow generated by a simulator, it is conceivable that the driving scenario to be verified will not be included in the random traffic flow even after the simulation.

[0016] Therefore, in one embodiment of the present invention, a simulation platform is configured that includes two simulators: a traffic flow simulator that generates random traffic flows and a verification simulator that verifies functionality and performance. The platform searches for driving scenarios to be verified in advance from the random traffic flows. Then, vehicle models that have been confirmed to be included in the driving scenarios to be verified are replaced with vehicle models for verification, and a simulation is performed.

[0017] "Configuration of driving scenario search device and simulation system" A driving scenario retrieval device and a simulation system according to an embodiment of the present invention will be described in detail below with reference to the drawings. Fig. 1 shows a schematic configuration of the simulation system according to the embodiment.

[0018] The simulation system 100 comprises a scenario database 10 that stores a scenario list including driving scenarios for verifying the functionality and performance of an autonomous driving function, a simulation execution management tool (hereinafter also referred to as an execution management tool or management device) 20 that manages the driving scenarios and performs operations from issuing instructions to executing a simulation to saving the results, and a simulation tool including a traffic flow simulator 30 and a verification simulator 40 that perform a simulation based on execution instructions from the execution management tool 20. The execution management tool 20, traffic flow simulator 30, and verification simulator 40 are connected to each other by signal lines.

[0019] The simulation system 100 is configured as a simulation platform for verifying the function and performance of a vehicle model equipped with an automatic driving function, and is configured to execute a cooperative simulation by linking a traffic flow simulator 30 and a verification simulator 40. In other words, the function and performance verification of the vehicle model equipped with an automatic driving function is executed as a cooperative simulation between the traffic flow simulator 30 and the verification simulator 40.

[0020] The scenario database 10 is a recording medium that stores multiple driving scenarios for verifying the functionality and performance of the automated driving function. Here, a driving scenario is a time-series representation of the behavior of traffic participants on a road network, and various driving scenarios can be generated depending on the interrelationships between traffic participants. Each driving scenario is given a scenario name and a scenario ID, and is stored in the scenario database 10 as a list. The scenario database 10 also stores a reference scenario pattern associated with each driving scenario for determining whether the driving scenario matches, as described below.

[0021] Figures 2(a) to 2(d) show examples of driving scenarios that can be the subject of verification. Figures 2(a) to 2(d) respectively show an overview of an overtaking scenario, a other vehicle cut-in scenario, a other vehicle cut-out scenario, and a following scenario. For example, the overtaking scenario shown in Figure 2(a) is a situation in which a rear vehicle A overtakes a front vehicle B on a straight road with two lanes on each side, and is given the scenario name "Straight road, two lanes on each side, overtaking" and a scenario ID of ID-1.

[0022] For example, the overtaking scenario is described as follows, where the vehicle behind is defined as the vehicle itself, the vehicle ahead is defined as the preceding vehicle, the lane on the left is defined as lane 1, and the lane on the right is defined as lane 2: (i) Stay in lane 1 → (ii) Follow the preceding vehicle → (iii) Change lane to the right → (iv) Stay in lane 2 → (v) Change lane to the left → (vi) Stay in lane 1. Note that here, an example is explained in which vehicles keep to the left.

[0023] The other vehicle cut-in scenario shown in Figure 2(b) is a situation in which another vehicle C cuts in from the right adjacent lane in front of the host vehicle D, which is traveling in the left adjacent lane on a straight road with two lanes on each side.The scenario name is ``Straight road, two lanes on each side, cut-in'' and the scenario ID is ID-2.

[0024] The other vehicle cutout scenario shown in Figure 2(c) is a situation in which another vehicle E ahead of the host vehicle F traveling in the adjacent lane to the left on a straight road with two lanes on each side cuts out. In the example shown in Figure 2(c), there is another preceding vehicle E0 ahead of the other vehicle E that is cutting out. The scenario name is "Straight road, two lanes on each side, cutout" and the scenario ID is ID-3.

[0025] The following scenario shown in Figure 2(d) is a situation in which the subject vehicle G follows another vehicle H traveling ahead on a straight road with two lanes on each side to keep in lane, and the scenario name is "Straight road, two lanes on each side, following" and the scenario ID is ID-4. As with the overtaking scenario described above, the other vehicle cut-in scenario, other vehicle cut-out scenario, and following scenario also describe the time-series transition of the surrounding environment including the subject vehicle and other vehicles.

[0026] The traffic flow simulator 30 is a simulation tool configured to set up a surrounding environment model including a road network and to simulate traffic flow on the road network. The traffic flow simulator 30 is configured, for example, by a computer, that is, a ROM that stores programs and data, a CPU that performs arithmetic processing, a RAM that reads out the programs and data and stores dynamic data and arithmetic processing results, and an input / output interface.

[0027] The traffic flow simulator 30 generates random traffic flow based on a traffic flow model by executing a program stored in ROM. Specifically, the behavior of traffic participants such as vehicles and pedestrians is generated as a traffic flow based on a traffic engineering model that theoretically represents the flow of vehicles and human behavior. As described above, the traffic flow generated by the traffic flow simulator 30 is a random traffic flow in which each vehicle model runs randomly while maintaining mutual relationships with other vehicles within the ranges of set parameters including vehicle speed, forward and rearward visibility distance, inter-vehicle distance, deceleration, etc.

[0028] The verification simulator 40 is a simulation tool configured to generate a vehicle model equipped with an automatic driving function and simulate driving the vehicle in order to verify the functionality and performance of the automatic driving function. The verification simulator 40 sets up a surrounding environment model including a sensor model, a vehicle control model, a vehicle model, and a road network, and performs a simulation in cooperation with the traffic flow simulator 30.

[0029] The verification simulator 40 is composed of, for example, a computer, that is, a ROM that stores programs and data, a CPU that performs arithmetic processing, a RAM that reads out the programs and data and stores dynamic data and arithmetic processing results, an input / output interface, etc. The verification simulator 40 executes the programs stored in the ROM to simulate the running of a vehicle model equipped with an automatic driving function in the traffic flow generated by the traffic flow simulator 30.

[0030] The execution management tool 20 is a controller that controls the traffic flow simulator 30 and the verification simulator 40 to execute and manage the linked simulation. The execution management tool 20 is composed of a computer, that is, a ROM that stores programs and data, a CPU that performs calculation processing, a RAM that reads the programs and data and stores dynamic data and calculation processing results, an input / output interface, etc., and performs operations from issuing instructions to execute a simulation to saving the results, and controls the entire simulation system 100.

[0031] The execution management tool 20 includes a driving scenario management unit 21, a traffic flow management unit 22 that controls the traffic flow simulator 30, a collaboration management unit 23 that controls a collaborative simulation between the traffic flow simulator 30 and the verification simulator 40, and a function verification unit 24 that verifies the functions and performance of a vehicle model equipped with an autonomous driving function based on the results of the collaborative simulation.

[0032] The driving scenario management unit 21 has a list of driving scenarios similar to the list stored in the scenario database 10. When a driving scenario to be verified is selected by a user via an input member (not shown), for example, the driving scenario management unit 21 obtains the selected specific driving scenario from the scenario database 10.

[0033] The traffic flow management unit 22 includes a setting instruction unit 221, a traffic flow log data generation unit 222, and a traffic flow log data analysis unit 223. The setting instruction unit 221 instructs the traffic flow simulator 30 to set a surrounding environment model and generate traffic flows, and causes the traffic flow simulator 30 to execute a simulation.

[0034] The simulation results of the traffic flow simulator 30 are input to the traffic flow management unit 22. The traffic flow log data generation unit 222 generates traffic flow log data as time-series data representing the behavior of each vehicle model included in the traffic flow generated by the traffic flow simulator 30. The traffic flow log data analysis unit 223 extracts and analyzes scenario patterns included in the traffic flow based on the generated traffic flow log data, and determines whether a specific driving scenario is included in the traffic flow.

[0035] The traffic flow management unit 22 and the traffic flow simulator 30 work together to function as a travel scenario search device that searches for a specific travel scenario. Details of how the travel scenario search device searches for a specific travel scenario will be described later.

[0036] The cooperation management unit 23 includes a setting instruction unit 231, a verification log data generation unit 232, and a verification log data analysis unit 233. The setting instruction unit 231 instructs the traffic flow simulator 30 and the verification simulator 40 to set a surrounding environment model, and also instructs the verification simulator 40 to set a vehicle model equipped with an autonomous driving function to be verified (hereinafter also referred to as a verification vehicle model), and a sensor model and a vehicle control model for the verification vehicle model. Note that the road network included in the surrounding environment model is common to both the traffic flow simulator 30 and the verification simulator 40.

[0037] The setting instruction unit 231 further specifies a vehicle model included in a specific driving scenario in the traffic flow generated by the traffic flow simulator 30, i.e., a vehicle model that is confirmed to execute a specific driving scenario (hereinafter also referred to as a specific vehicle model), replaces it with the verification vehicle model, and instructs the traffic flow simulator 30 and the verification simulator 40 to execute a collaborative simulation.

[0038] The results of the collaborative simulation are input to the collaborative management unit 23. The verification log data generation unit 232 generates verification log data as time-series data representing the behavior of the verification vehicle model traveling simulated in the traffic flow generated by the traffic flow simulator 30. The verification log data analysis unit 233 analyzes the behavior of the verification vehicle model based on the generated verification log data, and determines whether the verification vehicle model behaved in accordance with a specific driving scenario.

[0039] The function verification unit 24 verifies the functionality and performance of the autonomous driving function in a specific driving scenario based on the results of the collaborative simulation. The log data, simulation results, and verification results of the functionality and performance generated by the execution management tool 20 are stored in, for example, a memory (not shown).

[0040] "Operation of the driving scenario search device and simulation system" The flow of processing executed by the simulation system 100 according to this embodiment will be described with reference to the flowchart in Fig. 3. The simulation system 100 performs the following processes: (1) designation of a specific driving scenario to be verified (S1), (2) setting conditions for a traffic flow simulation (S2), (3) generation of traffic flow (S3), (4) generation of traffic flow log data (S4), (5) analysis of the traffic flow log data (S5), (6) determination of a match with the specific driving scenario (S6), (7) determination of a replacement vehicle (S7), (8) setting conditions for the linked simulation (S8), (9) execution of the linked simulation (S9), (10) generation of verification log data (S10), (11) behavior analysis of the verification vehicle model (S11), (12) determination of execution of the specific driving scenario (S12), and (13) verification of function and performance (S13).

[0041] The entire processing in the simulation 100 is controlled by the execution management tool 20, but in particular, the processing of S2 to S6 above relates to the search for a specific driving scenario controlled by the traffic flow management unit 22, and the processing of S7 to S12 above relates to the execution determination of a specific driving scenario controlled by the cooperation management unit 23. Each processing will be explained below.

[0042] (1) Specify the specific driving scenario you want to verify (S1) First, a specific driving scenario for which the functionality and performance of the autonomous driving function is to be verified is specified. For example, a user selects the name or scenario ID of the specific driving scenario to be verified from a list of driving scenarios displayed on a display (not shown) via an input member (not shown). The driving scenario management unit 21 acquires the selected specific driving scenario and a reference scenario pattern for determining whether the driving scenario matches, which will be described later, from the scenario database 10.

[0043] (2) Setting conditions for traffic flow simulation (S2) The setting instruction unit 221 of the traffic flow management unit 22 sets the ambient environment conditions for generating a traffic flow by the traffic flow simulator 30. The ambient environment conditions for generating a traffic flow are selected by the user, for example, through the operation of an input member (not shown).

[0044] Specifically, the road network, traffic signal status, traffic flow simulation execution time, and traffic flow model status are set. The road network includes, for example, the type of road on which traffic flow is generated, the number of lanes, and the presence or absence of intersections, and conditions related to the road structure are set according to the specific driving scenario selected in S1. The traffic signal status includes, for example, the lighting pattern of traffic signals present on the road network.

[0045] The execution time of the traffic flow simulation is the length of time to run the traffic flow simulation, and can be set to, for example, about one minute, taking into account the length of time required to run a specific driving scenario.The traffic flow model includes parameters such as the number of vehicles appearing per hour, departure points (locations) and arrival points (locations), initial vehicle speed, front and rear visibility distances, inter-vehicle distance, and deceleration, as well as their ranges.

[0046] (3) Traffic flow generation (S3) The setting instruction unit 221 of the traffic flow management unit 22 sends a simulation execution instruction to the traffic flow simulator 30 so that the traffic flow is generated under the ambient environmental conditions set in S2. Each vehicle model in the traffic flow runs within the set parameter range while maintaining its interrelationship with other vehicle models. As a result, the traffic flow simulator 30 generates a random micro traffic flow with each vehicle as a particle.

[0047] (4) Traffic flow log data generation (S4) The traffic flow log data generation unit 222 of the traffic flow management unit 22 acquires data on the behavior of each vehicle model included in the traffic flow simulated and generated in S3 from the traffic flow simulator 30, and generates traffic flow log data that represents the behavior of each vehicle model. The traffic flow log data is time-series data that represents the behavior history of the vehicle models generated by the traffic flow simulator 30, and is generated for each vehicle model included in the traffic flow. The traffic flow log data includes, for example, time-series position coordinates of the vehicle model, with the horizontal direction of the lane in which the vehicle model is traveling as the x-axis, the vertical direction as the y-axis, and time as the z-axis, as well as timestamps.

[0048] (5) Traffic flow log data analysis (S5) The traffic flow log data analysis unit 223 of the traffic flow management unit 22 extracts and analyzes scenario patterns based on the traffic flow log data generated in S4. First, the traffic flow log data analysis unit 223 creates a time-series driving trajectory of the vehicle model from the departure point to the arrival point on an xyz three-dimensional coordinate system using the time-series position coordinates included in the traffic flow log data of each vehicle model. Here, for example, the center of the rear wheel axle of the vehicle model is set as the reference point of the vehicle model, and the left edge of the left lane at the departure point is set as the origin (0,0) of the x-axis and y-axis, and the time-series position of the vehicle is plotted.

[0049] By overlaying the time-series driving trajectories generated for each vehicle model in the order of vehicle departure, it is possible to determine whether a traffic flow that matches a specific driving scenario has been generated. Specifically, by extracting the time-series driving trajectories of vehicle models that match a reference scenario pattern related to a specific driving scenario, and comparing the time-series driving trajectories that match the reference scenario pattern, it is determined whether a specific driving scenario is established.

[0050] Figures 4(a) to 4(d) show examples of reference scenario patterns for determining a specific driving scenario on a straight road with two lanes on each side. The reference scenario pattern represents the lateral and longitudinal movement characteristics of a vehicle associated with a specific driving scenario. Figures 4(a) to 4(d) schematically show a straight road with two lanes on each side, consisting of a left lane (lane 1) and a right lane (lane 2) defined by a center line CL and lane boundary line BL. The illustrated reference scenario pattern represents the time-series driving trajectory of a vehicle model from the starting point to the destination point, with the origin (0,0) at the left edge of lane 1 at the starting point of the driving simulation, the x-axis representing the lateral (width) direction of the lane, the y-axis representing the longitudinal direction, and the z-axis representing time. Note that the time-series driving trajectory also includes the concept of time, but for simplicity, Figures 4(a) to 4(d) only show x- and y-coordinates representing the lateral and longitudinal positions.

[0051] The first reference scenario pattern P1 shown in Figure 4(a) is described as follows: Keep in lane 1 → Change lane to the right → Keep in lane 2 → Change lane to the left → Keep in lane 1. The x-coordinate of the vehicle model in keeping in lane 1 is expressed as x≦a±w, where a is the distance from the left edge of lane 1 to the center of lane 1, and w is the allowable range of lateral sway when keeping in lane 1. A right lane change is a continuous coordinate movement in the +x-axis direction across the lane boundary line BL, and the x-coordinate of the vehicle model is expressed as ca≦v, where a≦x≦c. Here, c is the distance from the left edge of lane 1 to the center of lane 2, and v is the lateral movement distance equivalent to one lane. The x-coordinate of the vehicle model in keeping in lane 2 is expressed as x≦c±u, where u is the allowable range of lateral sway when keeping in lane 2. A left lane change is a continuous coordinate movement in the -x-axis direction across the lane boundary line BL, and the x-coordinate of the vehicle model is expressed as ca≦v, where a≦x≦c. The x-coordinate of the vehicle model in the first lane keeping is again x≦a±w.

[0052] The second reference scenario pattern P2 shown in Figure 4(b) is described as: Keep in second lane → Change lane to left → Keep in first lane. The x-coordinate of the vehicle model when keeping in second lane is expressed as x≦c±u. A left lane change is a continuous coordinate movement in the -x-axis direction that crosses the lane boundary line BL, and the x-coordinate of the vehicle model is expressed as ca≦v when a≦x≦c. The x-coordinate of the reference point of the vehicle model when keeping in first lane is expressed as x≦a±w.

[0053] The third reference scenario pattern P3 shown in Figure 4(c) is described as: Keep in the first lane → Change to the right lane → Keep in the second lane. The x-coordinate of the vehicle model when keeping in the first lane is expressed as x≦a±w. A right lane change is a continuous coordinate movement in the +x-axis direction across the lane boundary line BL, and the x-coordinate of the vehicle model is expressed as ca≦v when a≦x≦c. The x-coordinate of the vehicle model when keeping in the second lane is expressed as x≦c±u.

[0054] The fourth reference scenario pattern P4 shown in 4(d) is described as "keeping in the first lane." The x-coordinate of the vehicle model in keeping in the first lane is expressed as x≦a±w. Note that keeping in the second lane and other reference scenario patterns may also be included on a straight road with two lanes on each side, but are not discussed here.

[0055] The traffic flow log data analysis unit 223 analyzes the time-series travel trajectories of vehicle models included in the traffic flow and extracts multiple vehicle models that correspond to any of the first to fourth reference scenario patterns P1 to P4. Then, based on the reference scenario patterns and timestamps of the extracted multiple vehicle models, it searches for vehicle models that have time-series travel trajectories that form a specific travel scenario. A method for searching for vehicle models that have time-series travel trajectories that match a specific travel scenario will be described using Figures 5(a) to 5(d).

[0056] (A) Overtaking scenario As shown in Figure 5(a), the overtaking scenario is a combination of vehicle model A having a first reference scenario pattern P1 and vehicle model B having a fourth reference scenario pattern P4. Whether vehicle model A has overtaken vehicle model B or not depends on whether vehicle model A overlaps with vehicle model B in time while traveling from the departure point to the destination point, and vehicle model B must be traveling in lane 1 in overtaking section L1 where vehicle model A changes lanes to the right, stays in lane 2, and then changes lanes to the left.

[0057] First, vehicle model A having the first reference scenario pattern P1 is extracted from the multiple vehicle models included in the traffic flow. A time range corresponding to the overtaking section L1 is assigned to the time-series driving trajectory of vehicle model A. The assigned time range is the z-axis coordinate range corresponding to the start and end of the overtaking section L1 and the y-axis coordinate range corresponding to the longitudinal range of the lane in the overtaking section L1. Next, vehicle model B, which traveled in the time range corresponding to the overtaking section L1 according to the fourth reference scenario pattern P4, is extracted from the multiple vehicle models included in the traffic flow. Once vehicle model A and vehicle model B are extracted, it can be determined that the overtaking scenario shown in Figure 2(a) has been formed, and further, it can be determined that vehicle model A matches the scenario conditions of the overtaking scenario and has executed the overtaking scenario.

[0058] It is possible that there are multiple vehicle models A and multiple vehicle models B, or that there is no vehicle model A or B at all.

[0059] (B) Other vehicle cut-in scenario As shown in Figure 5(b), the other vehicle cut-in scenario is composed of a combination of a vehicle model C having a second reference scenario pattern P2 and a vehicle model D having a fourth reference scenario pattern P4. Whether vehicle model C has cut in before vehicle model D, which is a rear vehicle, depends on whether vehicle model C overlaps with vehicle model D in time while vehicle model C travels from the departure point to the arrival point, and vehicle model D must be traveling in lane 1 during cut-in section L2 where vehicle model C maintains lane 2, changes lanes to the left, and maintains lane 1.

[0060] First, vehicle model C having the second reference scenario pattern P2 is extracted from the multiple vehicle models included in the traffic flow. A time range corresponding to cut-in section L2 is assigned to the time-series driving trajectory of vehicle model C. The assigned time range is the z-axis coordinate range corresponding to the start and end of cut-in section L2 and the y-axis coordinate range corresponding to the longitudinal range of the lane in cut-in section L2. Next, vehicle model D, which traveled in the time range corresponding to cut-in section L2 according to the fourth reference scenario pattern P4, is extracted from the multiple vehicle models included in the traffic flow. Once vehicle model C and vehicle model D are extracted, it can be determined that the other vehicle cut-in scenario shown in Figure 2(b) has been formed, and further, it can be determined that vehicle model D matches the scenario conditions of the other vehicle cut-in scenario and has executed the other vehicle cut-in scenario.

[0061] It should be noted that there may be a plurality of vehicle models C and a plurality of vehicle models D, or there may be no vehicle models at all.

[0062] (C) Other vehicle cutout scenario As shown in Fig. 5(c), the other vehicle cutout scenario is a combination of a vehicle model E having a third reference scenario pattern P3 and a vehicle model F having a fourth reference scenario pattern P4. Whether or not vehicle model E has cut out in front of the following vehicle model F depends on whether vehicle model F is traveling in lane 1 during the time from the departure point to the arrival point, where vehicle model E overlaps with vehicle model F in time, and vehicle model E performs the following sequence: maintain in lane 1, change lanes to the right, maintain in lane 2.

[0063] First, vehicle model E having the third reference scenario pattern P3 is extracted from the multiple vehicle models included in the traffic flow. A time range corresponding to cutout section L3 is assigned to vehicle model E's time-series driving trajectory. The assigned time range is the z-axis coordinate range corresponding to the start and end of cutout section L3 and the y-axis coordinate range corresponding to the longitudinal range of the lane in cutout section L3. Next, vehicle model F, which traveled in the time range corresponding to cutout section L3 according to the fourth reference scenario pattern P4, is extracted from the multiple vehicle models included in the traffic flow. Once vehicle model E and vehicle model F are extracted, it can be determined that the other vehicle cutout scenario shown in Figure 2(c) has been formed, and further, it can be determined that vehicle model F matches the scenario conditions of the other vehicle cutout scenario and has executed the other vehicle cutout scenario.

[0064] It should be noted that there may be a plurality of vehicle models E and vehicle models F, or there may be no vehicle models E or F at all.

[0065] (D) Follow-up scenario As shown in Fig. 5(d), the following scenario is composed of a combination of vehicle model G and vehicle model H having a fourth reference scenario pattern P4. Whether vehicle model G is following vehicle model H, which is a preceding vehicle, or not depends on whether vehicle model H is traveling in the first lane during the period from the departure point to the arrival point, where vehicle model G overlaps with vehicle model H in time and vehicle model G is keeping in the first lane during the following section L4.

[0066] Therefore, first, vehicle model G having the fourth reference scenario pattern P4 is extracted from the multiple vehicle models included in the traffic flow. A time range corresponding to the following section L4 is assigned to the time-series driving trajectory of vehicle model G. The assigned time range is the z-axis coordinate range corresponding to the start and end points of the following section L4 and the y-axis coordinate range corresponding to the longitudinal range of the lane in the following section L4. Here, the following section L4 can be the section from the departure point to the arrival point of vehicle model G. Then, vehicle model H that traveled ahead of vehicle model G in the fourth reference scenario pattern P4 is extracted from the multiple vehicle models included in the traffic flow during the time range corresponding to the following section L4. Whether vehicle model H traveled ahead of vehicle G can be determined, for example, by comparing the y-coordinates and z-coordinates of vehicle model G and vehicle model H during the time range corresponding to the following section L4. When vehicle model G and vehicle model H are extracted, it can be determined that the following scenario shown in Figure 2(d) has been formed, and further, it can be determined that vehicle model G meets the scenario conditions of the following scenario and has executed the following scenario.

[0067] It should be noted that there may be a plurality of vehicle models G and a plurality of vehicle models H. Also, after extracting vehicle model H, vehicle model G that follows vehicle model H may be extracted.

[0068] (6) Matching determination of specific driving scenarios (S6) Based on the analysis results of S5, the traffic flow log data analysis unit 223 determines whether the traffic flow generated by the traffic flow simulator 30 includes the specific driving scenario specified in S1. If the traffic flow does not include the specific driving scenario, the process returns to S3, where the traffic flow generation (S3), traffic flow log data generation (S4), and traffic flow log data analysis (S5) processes are performed to repeatedly search for the specific driving scenario. As described above, in random traffic flows, individual vehicle models travel within the set parameter ranges, so even if the ambient environmental conditions set in S2 are the same, the results of the traffic flow simulation generated in S3 will be different each time. On the other hand, if the traffic flow includes the specific driving scenario, the process proceeds to S7 to transition to the execution of the linked simulation.

[0069] (7) Deciding on replacement vehicle (S7) Based on the analysis results by the traffic flow log data analysis unit 223, the setting instruction unit 231 of the cooperation management unit 23 determines a replacement vehicle model to be used to replace the verification vehicle model equipped with an automatic driving function in the cooperation simulation.

[0070] Specifically, a specific vehicle model that matches the scenario conditions of a specific driving scenario included in the traffic flow generated by the traffic flow simulator 30 is set as a replacement vehicle model to replace the verification vehicle model. Specifically, in the case of an overtaking scenario, vehicle model A is set as the replacement vehicle model, in the case of a other vehicle cut-in scenario, vehicle model D is set as the replacement vehicle model, in the case of a other vehicle cut-out scenario, vehicle model F is set as the replacement vehicle model, and in the case of a following scenario, vehicle model G is set as the replacement vehicle model.

[0071] Note that there may be multiple specific vehicle models that match the scenario conditions of a specific driving scenario, and in such a case, the vehicle model that departed earliest in terms of time can be selected from the multiple specific vehicle models as the replacement vehicle model. Alternatively, the system can be configured so that the user can select a desired vehicle model from the multiple specific vehicle models as the replacement vehicle model.

[0072] (8) Setting conditions for the collaborative simulation (S8) The setting instruction unit 231 of the cooperation management unit 23 sets the ambient environment conditions and the verification vehicle model for executing the cooperation simulation. In the cooperation simulation, in order to reproduce the specific driving scenario executed by the replacement vehicle model set in S7, the ambient environment conditions for generating traffic flow by the traffic flow simulator 30 are set to be the same as the ambient environment conditions set in S2.

[0073] The setting instruction unit 231 sets a sensor model and a vehicle control model for the verification vehicle model of the verification simulator 40 that is to be replaced with the replacement vehicle model. The sensor model and the vehicle control model are models of the sensor group and vehicle control system that the verification vehicle model is equipped with to simulate the execution of an autonomous driving function. Therefore, the type, performance, number, etc. of the necessary sensors are set according to the autonomous driving function to be verified, and the content of control in the longitudinal and lateral directions of the vehicle is set, and the sensor model and the vehicle control model are constructed. The control system in the longitudinal and lateral directions of the vehicle includes, for example, an automated lane keeping system (ALKS). The sensor model and the vehicle control model are set by the user, for example, by operating input members not shown.

[0074] (9) Execution of collaborative simulation (S9) The setting instruction unit 231 of the cooperation management unit 23 sends an instruction to execute a cooperation simulation to the traffic flow simulator 30 and the cooperation simulator 40 so as to perform a driving simulation of the verification vehicle model using the ambient environmental conditions, sensor model, and vehicle control model set in S8. As a result, a cooperation simulation is executed in which the verification vehicle model equipped with the automatic driving function generated by the verification simulator 40 runs simulated in the traffic flow generated, i.e., reproduced, by the traffic flow simulator 30.

[0075] (10) Generation of verification log data (S10) The verification log data generation unit 232 of the collaboration management unit 23 acquires data on the behavior of the verification vehicle model and other vehicle models traveling in the traffic flow simulated in S9 from the traffic flow simulator 30 and the collaboration simulator 40, and generates traffic flow log data that represents the behavior of each vehicle model. The traffic flow log data is time-series data that represents the behavior history of the vehicle models traveling in the traffic flow generated by the traffic flow simulator 30, and is generated for each vehicle model included in the traffic flow, as in S4 described above. Of the traffic flow log data generated here, the log data of the verification vehicle model is also used to verify the functionality and performance of the autonomous driving function, and therefore will hereinafter also be referred to as verification log data.

[0076] (11) Behavior analysis of the verification vehicle model (S11) The verification log data analysis unit 233 of the cooperation management unit 23 extracts and analyzes scenario patterns based on the traffic flow log data generated in S10, similar to S5 described above. First, the verification log data analysis unit 233 creates a time-series driving trajectory of the verification vehicle model from the departure point to the arrival point on an xyz three-dimensional coordinate system, using the time-series position coordinates included in the verification log data of the verification vehicle model. The verification log data analysis unit 233 further creates time-series driving trajectories, based on the traffic flow log data, for other vehicle models present in the traffic flow between the verification vehicle model's departure point and the arrival point.

[0077] The verification log data analysis unit 233 first analyzes the time-series driving trajectory of the verification vehicle model to extract a scenario pattern and determines which of the first to fourth reference scenario patterns P1 to P4 described above it corresponds to. Furthermore, it analyzes the time-series driving trajectory of the other vehicle model to extract a scenario pattern that corresponds to one of the first to fourth reference scenario patterns P1 to P4. The verification log data analysis unit 233 compares the extracted scenario pattern of the verification vehicle model with the scenario patterns of the other vehicle models to analyze whether the scenario pattern of the verification vehicle model matches the scenario conditions of a specific driving scenario.

[0078] (12) Determining whether to execute a specific driving scenario (S12) Based on the analysis results of S11, the verification log data analysis 233 determines whether the verification vehicle model meets the scenario conditions of the specific driving scenario specified in S1. If it is determined that the verification vehicle model does not meet the scenario conditions of the specific driving scenario, that is, that the verification vehicle model does not behave according to the specific driving scenario to be verified, the process returns to S9 and the collaborative simulation is executed again, generating verification log data (S10) and analyzing the behavior of the verification vehicle model (S11). As described above, in random traffic flow, individual vehicle models travel within the set parameter ranges, so the results of the collaborative simulation executed in S9 will be different each time, even if the ambient environmental conditions and vehicle model settings for the collaborative simulation set in S8 are the same.

[0079] If it is determined that the verification vehicle model meets the scenario conditions of the specific driving scenario, that is, that the verification vehicle model behaves in accordance with the specific driving scenario to be verified, the process proceeds to S13.

[0080] (13) Verification of function and performance (S13) If it is determined that the verification vehicle model behaves in accordance with a specific driving scenario, the function verification unit 24 verifies the functionality and performance of the autonomous driving function in the specific driving scenario based on the verification log data of the verification vehicle model. Note that the function and performance verification process by the function verification unit 24 is performed optionally and can be omitted.

[0081] The traveling scenario searching device and simulation system 100 according to the present embodiment described above can achieve the following advantageous effects.

[0082] (1) The driving scenario search device includes a traffic flow simulator 30 that simulates traffic flow on a road network, a traffic flow log data generation unit 222 that generates time-series data representing the behavior of each vehicle model included in the traffic flow generated by the traffic flow simulator 30 as traffic flow log data, and a traffic flow log data analysis unit 223 that analyzes the traffic flow log data generated by the traffic flow log data generation unit 222 and searches for a specific vehicle model that matches a specific driving scenario from the vehicle models included in the traffic flow.

[0083] By searching for a specific vehicle model that matches a specific driving scenario from among the vehicle models included in the traffic flow generated by the traffic flow simulator, it is possible to determine whether the specific driving scenario to be verified has been executed and to identify the target vehicle for verification of the specific driving scenario.This makes it possible to confirm whether the specific driving scenario is included in the traffic flow before verifying the specific driving scenario through simulation.In addition, since it is possible to identify a specific vehicle model that matches the specific driving scenario, it is possible to specify a vehicle model to be replaced for verification.

[0084] (2) When there are multiple specific vehicle models that match a specific driving scenario, the traffic flow log data analysis unit 223 selects the vehicle model that departed earliest in time series as the specific vehicle model. This allows for quick search for the specific vehicle model, and enables early execution of simulations for verification.

[0085] (3) The traffic flow log data analysis unit 223 creates a time-series driving trajectory of each vehicle included in the traffic flow and compares the time-series driving trajectories that match the reference scenario pattern associated with the specific driving scenario to determine whether the specific driving scenario is established. This makes it possible to determine whether the specific driving scenario is established and to identify vehicle models that match the specific driving scenario. Note that since time-series driving trajectories are compared, it is possible to compare the driving trajectories of vehicle models that travel within a specific time range. Therefore, even if a vehicle model's driving trajectory matches the reference scenario pattern, a vehicle model that travels within a different time range is excluded from the determination of whether the specific driving scenario is established, allowing for a more detailed determination.

[0086] (4) The traffic flow simulator 30 generates traffic flows in which each vehicle model behaves randomly within a set parameter range, so that different traffic flows can be generated without changing the vehicle model parameters for each simulation.

[0087] (5) The simulation system 100 is configured to perform theoretical verification of the autonomous driving function of a vehicle, and includes the above-mentioned driving scenario search device, a verification simulator 40 that generates a verification vehicle model that simulates the autonomous driving function, and a simulation execution management tool (management device) 20 that operates the verification simulator 40 in conjunction with the traffic flow simulator 30. The execution management tool 20 designates a specific vehicle model searched by the driving scenario search device as a replacement vehicle model, and replaces the verification vehicle model of the verification simulator 40 with the replacement vehicle model in the traffic flow generated by the traffic flow simulator 30, and simulates the vehicle running.

[0088] A specific vehicle model that is confirmed to be included in a specific driving scenario to be verified is designated as a replacement vehicle model and is configured to replace the verification vehicle model, so that it is possible to avoid a situation in which the specific driving scenario to be verified cannot be found in the cooperative simulation of the traffic flow simulator 30 and the verification simulator 40.

[0089] (6) The execution management tool 20 has a verification log data generation unit 232 that generates time-series data representing the behavior of the verification vehicle model obtained from the verification simulator 40 as verification log data, and a verification log data analysis unit 233 that analyzes the verification log data generated by the verification log data generation unit 232 and determines whether the verification vehicle model behaved according to a specific driving scenario. This makes it possible to confirm whether a specific driving scenario to be verified was executed in the cooperative simulation of the traffic flow simulator 30 and the verification simulator 40.

[0090] (7) If the verification log data analysis unit 233 determines that the verification vehicle model did not behave according to the specific driving scenario, the execution management tool 20 replaces the verification vehicle model with a replacement vehicle model and runs it again in a simulated manner. This allows the specific driving scenario to be verified to be executed reliably.

[0091] (8) The execution management tool 20 further includes a function verification unit 24 that verifies the autonomous driving function in a specific driving scenario based on the verification log data of the verification vehicle model when the verification log data analysis unit 233 determines that the verification vehicle model behaved in accordance with the specific driving scenario. This makes it possible to reliably verify the autonomous driving function in the specific driving scenario that is to be verified.

[0092] -Variations- (1) In the above-described embodiment, a micro traffic flow based on a traffic engineering model is generated by the traffic flow simulator 30. Here, as a model of an individual vehicle set as a particle of the micro traffic flow, a model also called a C and C Driver Model (Capable and Careful Driver Model) can be used, in which parameters are set to be able to simulate the driving of a test driver, an experienced driver, or a professional driver such as a taxi driver.

[0093] (2) The individual vehicle models set as particles of the microscopic traffic flow may be vehicle models equipped with an automatic driving function. By using the traffic flow simulator 30 to model all vehicles included in the traffic flow generated by the traffic flow simulator 30 as vehicle models equipped with an automatic driving function, it is possible to perform a simulation, for example, assuming a dedicated road on which only vehicles equipped with an automatic driving function can travel.

[0094] (3) In the above-described embodiment, the simulation system 100 is configured to include the scenario database 10, the simulation execution management tool 20, the traffic flow simulator 30, and the verification simulator 40. However, the configuration of the simulation system 100 is not limited to this. For example, the simulation execution management tool 20 and the verification simulator 40 may be configured as an integrated unit, or the driving scenario database 10 may be incorporated into the simulation execution management tool 20. Also, the function verification unit 24 of the simulation execution management tool 20 may be omitted, and the simulation system 100 may be configured to execute up to the execution determination (S12) of a specific driving scenario.

[0095] (4) A driving scenario search device including only the traffic flow management unit 22 of the simulation execution management tool 20 and the traffic flow simulator 30 can be configured independently of the simulation system 100. In this case, the driving scenario search device can execute the processes from S1 to S6 in the flowchart of Fig. 3 and can be configured to store the search results for each specific driving scenario in a memory (not shown). The specific driving scenarios searched and stored by the driving scenario search device can be used in the linked simulation executed by the simulation system 100.

[0096] (5) In the above-described embodiment, specific driving scenarios include an overtaking scenario, a vehicle cut-in scenario, a vehicle cut-out scenario, and a following scenario on a straight road with two lanes on each side. However, driving scenarios for verifying the functionality and performance of the autonomous driving function are not limited to these, and any scenario can be envisioned. For example, road structures include curves with two lanes on each side, straight roads with three lanes on each side, and intersections. Furthermore, the behavior of the verification vehicle model also includes, for example, following a preceding vehicle on a curve and turning right or left at an intersection. Furthermore, scenarios that combine vehicle cut-out and following are also possible, such as when vehicle E0 is further ahead after vehicle E cuts out, as shown in FIG. 2(c).

[0097] (6) The processing flow in the simulation system 100 is not limited to that shown in the flowchart of Fig. 3. For example, the condition setting processes in S2 and S8 may be set in advance for each driving scenario.

[0098] Although several embodiments of the present invention have been described above, it should be noted that the present invention is not limited to the above-described embodiments, and various modifications and variations are possible within the scope of the present invention. [Explanation of symbols]

[0099] 20 Simulation execution management tool (management device) 22 Traffic Flow Management Department 222 Traffic flow log data generation unit 223 Traffic Flow Log Data Analysis Department 23 Cooperation Management Department 232 Verification log data generation unit 233 Verification Log Data Analysis Unit 24 Functionality Verification Department 30 Traffic flow simulator 40 Verification Simulator 100 Simulation Systems

Claims

1. a traffic flow simulator that simulates traffic flow on a road network; a traffic flow log data generation unit that generates time-series data representing the behavior of each vehicle model included in the traffic flow generated by the traffic flow simulator as traffic flow log data; a traffic flow log data analysis unit that analyzes the traffic flow log data generated by the traffic flow log data generation unit and searches for a specific vehicle model that matches a specific driving scenario from vehicle models included in the traffic flow; A driving scenario search device comprising:

2. 2. The traveling scenario search device according to claim 1, wherein, when there are multiple specific vehicle models that match the specific traveling scenario, the traffic flow log data analysis unit selects, as the specific vehicle model, the vehicle model that departed at the earliest timing in a chronological order among the multiple specific vehicle models.

3. 3. The driving scenario search device according to claim 1, wherein the traffic flow log data analysis unit creates a time-series driving trajectory of each vehicle included in the traffic flow, and determines whether the specific driving scenario is established by comparing the time-series driving trajectories that match a reference scenario pattern related to the specific driving scenario.

4. 3. The driving scenario search device according to claim 1, wherein the traffic flow simulator generates a traffic flow in which each vehicle model behaves randomly within a set parameter range.

5. A simulation system for desktop verification of an automatic driving function of a vehicle, The driving scenario search device according to claim 1 ; a verification simulator that generates a verification vehicle model that simulates autonomous driving functions; a management device that operates the verification simulator and the traffic flow simulator in cooperation with each other; Equipped with The management device designates the specific vehicle model searched for by the driving scenario search device as a replacement vehicle model, and in the traffic flow generated by the traffic flow simulator, replaces the verification vehicle model of the verification simulator with the replacement vehicle model and simulates a traffic flow.

6. The management device a verification log data generation unit that generates time-series data representing a behavior of the verification vehicle model obtained from the verification simulator as verification log data; a verification log data analysis unit that analyzes the verification log data generated by the verification log data generation unit and determines whether the verification vehicle model behaves in accordance with the specific driving scenario; The simulation system of claim 5 , comprising:

7. The simulation system of claim 6, wherein if the verification log data analysis unit determines that the verification vehicle model did not behave in accordance with the specific driving scenario, the management device replaces the verification vehicle model with the replacement vehicle model and simulates driving it again.

8. The simulation system of claim 6 or 7, wherein the management device further has a function verification unit that verifies the autonomous driving function in the specific driving scenario based on the verification log data of the verification vehicle model when the verification log data analysis unit determines that the verification vehicle model has behaved in accordance with the specific driving scenario.

Images