Scenario parameter optimization device, scenario parameter optimization method, scenario parameter optimization program, and control logic inspection system

Abstract

The scenario parameter optimization device includes an objective function value calculation unit (11) that calculates an objective function value of a predetermined event and an event occurrence time at which the predetermined event occurs based on the execution result of a simulation test of a scenario including a first agent and a second agent, which is executed based on scenario parameters; a responsibility determination unit (12) that determines whether or not the first agent is responsible for the occurrence of the predetermined event from the situation at a responsibility determination time prior to the event occurrence time, based on the execution result and the event occurrence time, and outputs the determination result as a responsibility determination result; and a scenario parameter optimization unit (13) that optimizes the scenario parameters based on the objective function value and the responsibility determination result so that the objective function value becomes smaller and it is determined that the first agent is responsible for the occurrence of the predetermined event.

Description

[Technical field]

[0001] The present disclosure relates to scenario parameter optimization techniques. [Background technology]

[0002] Technologies have been developed that allow agents such as cars or robots to drive autonomously. To allow an agent to drive autonomously, it is necessary to test the control logic for controlling the agent in various scenarios that simulate real environments. One method for testing control logic is called search-based testing. Search-based testing refers to a method of searching for a scenario to be tested using a metaheuristic optimization technique. Patent Document 1 is a document that discloses such a search-based testing technique. Patent Document 1 discloses a technique for setting parameters of a scenario used in a simulation test of a program to be tested by using an optimization process. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2018-190230 A Summary of the Invention [Problem to be solved by the invention]

[0004] The technology of Patent Document 1 does not anticipate scenarios in which multiple agents exist, and therefore has the problem that it is difficult to optimize scenario parameters used in a scenario in which multiple agents exist.

[0005] The present disclosure has been made to solve such problems, and aims to provide a scenario parameter optimization technique that can optimize scenario parameters used in a scenario in which multiple agents exist. [Means for solving the problem]

[0006] One aspect of a scenario parameter optimization device according to an embodiment of the present disclosure includes an objective function value calculation unit that accepts execution results of a simulation test of a scenario including a first agent and a second agent, which is executed based on scenario parameters, and calculates, based on the execution results, an objective function value of a predetermined event from an objective function expressing the predetermined event and an event occurrence time at which the predetermined event occurs; a responsibility determination unit that determines, based on the execution result and the calculated event occurrence time, whether the first agent is responsible for the occurrence of the predetermined event from a situation at a responsibility determination time, which is a time prior to the calculated event occurrence time, and outputs the result of the determination as a responsibility determination result; and a scenario parameter optimization unit that optimizes the scenario parameters based on the calculated objective function value and the output responsibility determination result, so that the calculated objective function value becomes smaller and it is determined that the first agent is responsible for the occurrence of the predetermined event. Effect of the Invention

[0007] A scenario parameter optimization unit according to an embodiment of the present disclosure can optimize scenario parameters used in a scenario in which multiple agents exist. [Brief description of the drawings]

[0008] [Figure 1] 1 is a functional block diagram showing a configuration example of a scenario parameter optimization device and a control logic inspection system according to a first embodiment. FIG. [Figure 2A] FIG. 2 is a diagram illustrating an example of the hardware configuration of a scenario parameter optimization device and a control logic inspection system. [Figure 2B] FIG. 2 is a diagram illustrating an example of the hardware configuration of a scenario parameter optimization device and a control logic inspection system. [Diagram 3] 1 is a diagram showing a part of a flowchart of a scenario parameter optimization method, more specifically, an operation by an objective function value calculation unit of a scenario parameter optimization device. [Figure 4] 1 is a diagram showing a part of a flowchart of a scenario parameter optimization method, more specifically, the operation of a responsibility determination unit of the scenario parameter optimization device. [Diagram 5] 1 is a diagram showing a part of a flowchart of a scenario parameter optimization method, more specifically, the operation of a responsibility determination unit of the scenario parameter optimization device. [Figure 6] 1 is a diagram showing a part of a flowchart of a scenario parameter optimization method, more specifically, an operation by a scenario parameter optimization unit of a scenario parameter optimization device. [Figure 7] Fig. 7A is a diagram illustrating the operation of the scenario parameter optimization device when the lateral distance is equal to or less than the vehicle width, and Fig. 7B is a diagram illustrating the operation of the scenario parameter optimization device when the lateral distance is greater than the vehicle width. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Various embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. Note that components with the same or similar reference numerals in the drawings have the same or similar configurations or functions, and redundant descriptions of such components will be omitted.

[0010] In this disclosure, the term "or" is used to mean an inclusive logical or unless otherwise specified. When the term "or" is used to mean an exclusive logical or, this will be clearly stated.

[0011] Embodiment 1 <Configuration> A scenario parameter optimization device and a control logic inspection system according to a first embodiment of the present disclosure will be described with reference to Fig. 1. Fig. 1 is a functional block diagram showing a configuration example of a scenario parameter optimization device 10 according to the first embodiment, and a control logic inspection system 1 including the scenario parameter optimization device 10. As shown in Fig. 1, the control logic inspection system 1 includes a scenario parameter optimization device 10 and a simulation execution device 20. The simulation execution device 20 is a device that performs inspection of a control logic program using a scenario. The scenario parameter optimization device 10 is a device that optimizes scenario parameters used in a scenario that the simulation execution device 20 uses when performing inspection. Each device will be described below.

[0012] (Simulation execution device) The simulation execution device 20 is a device that receives a scenario setting D1, values ​​of scenario parameters used in the scenario setting D1, and a control logic program D2 as inputs, and executes a simulation test in which the control logic program D2 operates under a scenario specifically set by the values ​​of the scenario parameters. The simulation execution device 20 executes the simulation test according to a calculation program and data for the simulation test (not shown). The calculation program and data (not shown) include a sensor model such as a radar or a camera.

[0013] The control logic specifies the behavior of the controlled agent, which is the object to be controlled. More specifically, the control logic controls the behavior of the controlled agent based on the result of the controlled agent's perception of the outside world. For example, when the control logic recognizes an object in the moving direction of the controlled agent, the control logic slows down the moving speed of the controlled agent or changes the moving direction of the controlled agent so as to avoid a collision with the recognized object. The amount of control is determined, for example, every 100 ms. The control logic program D2 is a program that realizes the control logic.

[0014] In the present disclosure, an agent that is the object of control by the control logic is referred to as a controlled agent, and an agent other than the controlled agent that exists around the controlled agent is referred to as a surrounding agent. In addition, in the present disclosure, the controlled agent and the surrounding agent may be more generally referred to as a first agent and a second agent, respectively. Examples of agents include any moving object such as a vehicle, a robot, or a drone.

[0015] A scenario means a time-dependent development between multiple scenes. A scene is a term that is determined at a certain time. When the agent is a vehicle, the scenario includes road conditions and a target position of the controlled vehicle. Examples of scenarios when the agent is a vehicle include, for example, 1) a scenario in which the controlled vehicle runs in a certain lane at a certain time, surrounding vehicles run 60 m ahead of the controlled vehicle in a lane adjacent to the lane, and the surrounding vehicles change lanes to a position 40 m ahead of the controlled vehicle within a period from that time to a specified time, or 2) a scenario in which the controlled vehicle runs in a certain lane at a certain time, surrounding vehicles run 20 m ahead of the controlled vehicle in a lane adjacent to the lane, and the controlled vehicle changes lanes to a position 40 m behind the surrounding vehicles within a period from that time to a specified time.

[0016] A setting for at least one scenario parameter is given to the scenario. Examples of the scenario parameters include the initial speed of the controlled agent, and the initial position, initial speed, target speed, and acceleration of the surrounding agents. If the agent is a vehicle, the scenario parameters may include a target lane. The scenario parameter optimization device 10 is given an initial value of the scenario parameter D9 by, for example, a user input, and outputs the initial value of the scenario parameter D9 to the simulation execution device 20. The simulation execution device 20 optimizes the value of the scenario parameter D9 based on the objective function value D4 and the responsibility determination result D8.

[0017] The simulation execution device 20 executes a simulation test on the behavior of the controlled agent by the control logic program D2 under a simulated scenario setting D1 specifically defined by scenario parameters. The simulation execution device 20 outputs a simulation execution result D3, which is the execution result of the simulation test, to the scenario parameter optimization device 10. The simulation execution result D3 includes a movement log that records the movements of the controlled agent and surrounding agents. The movement log includes, for example, the position, speed, and acceleration of each agent at each time during the simulation test, as well as the recognition state of the surrounding agents by the controlled agent.

[0018] (Scenario parameter optimization device) The scenario parameter optimization device 10 is a device for receiving the simulation execution result D3, analyzing the simulation execution result D3, and optimizing the scenario parameters used in the scenario setting D1. To realize such functions, the scenario parameter optimization device 10 includes an objective function value calculation unit 11, a responsibility determination unit 12, and a scenario parameter optimization unit 13. The responsibility determination unit 12 includes a responsibility determination time detection unit 121 and a responsibility attribution determination unit 122. Each functional unit will be described in detail below.

[0019] (Objective function value calculation part) The objective function value calculation unit 11 calculates an objective function value D4 of a predetermined event from an objective function expressing the predetermined event based on the simulation execution result D3, and outputs the calculated objective function value D4 to the scenario parameter optimization unit 13. The predetermined event is, for example, an event in which the inter-agent distance between the controlled agent and the surrounding agents becomes small over time, thereby minimizing the inter-agent distance. In addition, if the event in which the inter-agent distance becomes small is the predetermined event, the objective function expressing the predetermined event is a function expressing the distance between the agents. If the minimum value of the inter-agent distance is smaller than a certain threshold, it means that a problem of collision between agents will occur.

[0020] The calculation of the objective function value D4 will be specifically described with reference to FIG. 7. In FIG. 7A and FIG. 7B, the controlled agent is the controlled vehicle V1, and the surrounding agent is the surrounding vehicle V2. The centers of gravity of the controlled vehicle V1 and the surrounding vehicle V2 are (x1, y1) and (x2, y2), respectively. As in the case of FIG. 7A, when the distance |y2-y1| in the lateral direction y between the controlled vehicle V1 and the surrounding vehicle V2 is equal to or less than the vehicle width W, the objective function value calculation unit 11 calculates the minimum value, which is the distance at the closest approach in the longitudinal direction x, which is the traveling direction of the vehicle, as the objective function value D4. For example, a fixed value of 2000 mm may be used as the vehicle width W. When the vehicle width W1 of the controlled vehicle V1 is different from the vehicle width W2 of the surrounding vehicle V2, the average value of the vehicle width W1 and the vehicle width W2 may be used as the vehicle width W.

[0021] In contrast, when the lateral y distance |y2-y1| is greater than the vehicle width W, as in the case of Figure 7B, the control target vehicle V1 and the surrounding vehicle V2 will not collide, so the objective function value calculation unit 11 does not calculate the inter-vehicle distance at the closest point between the control target vehicle V1 and the surrounding vehicle V2, and regards the inter-vehicle distance as a value large enough not to affect the optimization result (for example, 10,000 m).

[0022] Moreover, the objective function value calculation unit 11 outputs the time when the data used to calculate the objective function value D4 was acquired as the event occurrence time D5 to the responsibility determination time detection unit 121. In other words, the event occurrence time D5 is the occurrence time of a predetermined event.

[0023] (Responsibility Determination Department) The responsibility determination unit 12 is a functional unit that determines whether the controlled agent is responsible for the occurrence of a predetermined event. More specifically, based on the simulation execution result D3 and the event occurrence time D5, it determines whether the controlled agent is responsible for the occurrence of a predetermined event from the situation at a responsibility determination time D6, which is a time before the event occurrence time D5, and outputs the determination result as a responsibility determination result D8. To realize such a function, the responsibility determination unit 12 includes a responsibility determination time detection unit 121 and a responsibility attribution determination unit 122.

[0024] (Responsibility determination time detection unit) Based on the simulation execution result D3, the responsibility judgment time detection unit 121 detects the time when a behavior of a surrounding agent occurred that leads to the occurrence of a predetermined event, which occurred before the event occurrence time D5, as a candidate responsibility judgment time, determines a responsibility judgment time D6 based on the detected candidate responsibility judgment time, and outputs the determined responsibility judgment time D6 to the responsibility attribution determination unit 122.

[0025] Here, an example of the behavior of the surrounding agent that leads to the occurrence of a predetermined event will be described. In order to describe the example, a vehicle is assumed as an agent, and a scenario is assumed in which two vehicles, a controlled vehicle and a surrounding vehicle, travel on a road. In addition, an event in which the inter-vehicle distance between the controlled vehicle and the surrounding vehicles becomes a minimum value over time is assumed as the predetermined event. Under such an assumption, examples of such behavior of the surrounding agent can include course change (lane change), acceleration, or deceleration.

[0026] More specifically, an example of lane change is when a surrounding vehicle traveling in an adjacent lane cuts in front of the controlled vehicle while the controlled vehicle is traveling in a lane and changes lanes. A scenario in which the surrounding vehicle decelerates before and after the lane change may be assumed.

[0027] Another example of lane change is when the controlled vehicle starts to change lanes from a first lane to a second lane adjacent to the first lane, and then a surrounding vehicle traveling in a third lane adjacent to the second lane changes lanes to the second lane. A scenario in which the surrounding vehicle decelerates before and after the lane change may be assumed.

[0028] An example of acceleration is when a controlled vehicle traveling in a certain lane starts to change lanes to an adjacent lane, and then a surrounding vehicle traveling in the adjacent lane behind the controlled vehicle accelerates.

[0029] An example of deceleration is when a surrounding vehicle ahead of the controlled vehicle traveling in the same lane suddenly decelerates.

[0030] In these examples, the behavior of the surrounding vehicles is such that the inter-vehicle distance between the controlled vehicle and the surrounding vehicles becomes shorter. The responsibility judgment time detection unit 121 detects the time when such behavior of the surrounding agents occurs as a responsibility judgment time candidate. When the surrounding agents perform multiple behaviors, such as decelerating after changing lanes, the responsibility judgment time detection unit 121 calculates the latest time before the event occurrence time D5 at which a change in the behavior of the surrounding agents occurs as a responsibility judgment time candidate. The responsibility judgment time detection unit 121 determines a responsibility judgment time D6 based on this responsibility judgment time candidate, and outputs the determined responsibility judgment time D6 to the responsibility attribution determination unit 122. The determination of the responsibility judgment time D6 based on the responsibility judgment time candidate will be described later.

[0031] (Responsibility Attribution Determination Department) The responsibility attribution determination unit 122 determines whether or not the controlled agent is responsible for the occurrence of a predetermined event based on the situation at the responsibility determination time D6 in the simulation execution result D3, and outputs the determination result as a responsibility determination result D8 to the scenario parameter optimization unit 13. The situation means the physical quantities of the controlled agent and surrounding agents. Examples of the physical quantities include position, speed, and acceleration. The method of determination will be described later.

[0032] (Scenario Parameter Optimization Department) The scenario parameter optimization unit 13 optimizes the scenario parameters D9 based on the objective function value D4 and the responsibility determination result D8 so that the controlled agent is responsible for the occurrence of a predetermined event and the objective function value D4 is small. Optimization methods include heuristic methods such as Hill Climbing, Simulated Annealing, and Genetic Algorithms.

[0033] Next, referring to Figures 2A and 2B, an example of the hardware configuration of the scenario parameter optimization device 10 and the simulation execution device 20 will be described. The functions of each device are realized by a processing circuitry. The processing circuitry may be a dedicated processing circuit 100a as shown in Figure 2A, or a computer including a memory 100c and a processor 100b as shown in Figure 2B. The processor 100b reads a program stored in the memory 100c and executes the read program.

[0034] When the processing circuitry is a dedicated processing circuit 100a, the dedicated processing circuit 100a may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination of these. The functions of each device may be realized by separate processing circuits, or multiple functions may be realized together in a single processing circuit.

[0035] When the processing circuitry is the processor 100b, the functions of each device are realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the memory 100c. The processor 100b realizes the functions of each device by reading and executing the programs stored in the memory. Here, examples of the memory 100c include non-volatile or volatile semiconductor memories such as random access memory (RAM), read-only memory (ROM), flash memory, erasable programmable read-only memory (EPROM), and electrically erasable programmable read-only memory (EEPROM), as well as magnetic disks, flexible disks, optical disks, compact disks, mini disks, and DVDs.

[0036] <Operation> Next, the operation of the scenario parameter optimization device 10 will be described with reference to Fig. 3 to Fig. 6. Fig. 3 shows the operation of the objective function value calculation unit 11 of the scenario parameter optimization device 10. Fig. 4 and Fig. 5 show the operation of the responsibility determination unit 12 of the scenario parameter optimization device 10. Fig. 6 shows the operation of the scenario parameter optimization unit 13 of the scenario parameter optimization device 10. A detailed explanation will be given below with reference to each figure.

[0037] The operation of the objective function value calculation unit 11 will be described with reference to FIG. 3. In step ST11, the objective function value calculation unit 11 analyzes the simulation execution result D3 output from the simulation execution device 20. The simulation execution result D3 includes a movement log such as a running log or a flight log in which position information of the controlled agent and at least one surrounding agent is recorded. When multiple surrounding agents exist, the simulation execution result D3 includes the movement log of each surrounding agent. The objective function value calculation unit 11 analyzes the simulation execution result D3 to find the minimum value of the distance from the controlled agent to at least one surrounding agent. When multiple surrounding agents exist, the objective function value calculation unit 11 finds the minimum value of the distance to each surrounding agent.

[0038] In step ST12, the objective function value calculation unit 11 calculates an objective function value from the minimum value of one or more distances to the obtained surrounding agents, and outputs the calculated objective function value D4. When only a single surrounding agent exists, the objective function value D4 is the minimum value of the distance to the surrounding agent. When multiple surrounding agents exist, the objective function value D4 may be multiple values, the number of which is equal to the number of surrounding agents, or may be a single value that is the minimum value of the distance to the controlled agent. The objective function value calculation unit 11 outputs the calculated objective function value D4 to the scenario parameter optimization unit 13.

[0039] In step ST13, the objective function value calculation unit 11 outputs the time when the data used to calculate the objective function value D4 was acquired, i.e., the time when the distance between the controlled agent and the surrounding agents is smallest, to the responsibility determination time detection unit 121 as the event occurrence time D5.

[0040] 4 and 5, the operation of the responsibility determination unit 12 including the responsibility determination time detection unit 121 and the responsibility attribution determination unit 122 will be described. In step ST21, the responsibility determination time detection unit 121 analyzes the simulation execution result D3 output from the simulation execution device 20, and detects the behavior change time at which a behavior change occurred in each surrounding agent. If the surrounding agent is, for example, a car, examples of the behavior change include lane changing, acceleration, deceleration, etc.

[0041] In step ST22, the responsibility determination time detection section 121 determines, for each surrounding agent, the latest behavior change time prior to the event occurrence time D5 among the behavior change times of each surrounding agent, as a responsibility determination time candidate.

[0042] In step ST23, the responsibility determination time detection unit 121 analyzes the simulation execution result D3, detects the recognition start time at which the controlled agent starts recognizing the surrounding agents before the event occurrence time D5 for each surrounding agent, calculates the later of the responsibility determination time candidate and the recognition start time as the responsibility determination time D6, and outputs the calculated responsibility determination time D6 to the responsibility attribution determination unit 122. For example, when a surrounding vehicle in front of the control target vehicle in an adjacent lane changes lanes and cuts in front of the control target vehicle, the recognition start time at which the controlled vehicle starts recognizing that the surrounding vehicle is entering the same lane as the controlled vehicle may be later than the responsibility determination time candidate at which the surrounding vehicle starts changing lanes. In such an example, the responsibility determination time detection unit 121 calculates the later recognition start time as the responsibility determination time D6.

[0043] In step ST24, the responsibility attribution determination section 122 extracts data of the surrounding agents at the responsibility determination time D6 calculated by the responsibility determination time detection section 121 from the simulation execution result D3 output from the simulation execution device 20.

[0044] In step ST25, the responsibility attribution determination unit 122 determines whether the controlled agent is responsible in relation to each surrounding agent for the occurrence of a predetermined event, and outputs the determination result to the scenario parameter optimization unit 13 as a responsibility determination result D8.

[0045] FIG. 5 is a flow chart for explaining in detail the process by the responsibility attribution determination unit 122 in step ST25. In step ST251, based on the data of the surrounding agents at the responsibility determination time D6, it is determined whether or not the controlled agent was able to avoid the occurrence of the predetermined event with a realistic control amount during the period from the responsibility determination time D6 to the event occurrence time D5. For example, one or both of the deceleration value and the turning angular velocity value required for collision avoidance are calculated according to the law of motion, and if the calculated value exceeds a predetermined threshold value D7, it is determined that the occurrence of the predetermined event cannot be avoided with a realistic control amount (NO in step ST251). In the case of deceleration, for example, a value of about 0.75 G, which is the braking force that an experienced driver can generate, or a value of about 0.85 G, which is the braking force when a collision damage mitigation brake device is activated, can be used as the threshold value D7. On the other hand, if the calculated value is equal to or less than the predetermined threshold value D7, it is determined that the occurrence of the predetermined event can be avoided with a realistic control amount (YES in step ST251).

[0046] If the answer is YES in step ST251, the process proceeds to step ST252, where it is determined that the controlled agent is responsible.

[0047] If the answer is NO in step ST251, the process proceeds to step ST253, where it is determined that the controlled agent is not responsible.

[0048] In step ST254, a behavior change of the controlled agent is detected from the simulation execution result D3 output from the simulation execution device 20, and if the time of the detected behavior change is later than the responsibility determination time D6, the result of the determination is corrected so that the controlled agent is determined to be responsible. For example, if the surrounding vehicle ahead in the adjacent lane decelerates and then the controlled vehicle changes lanes into the adjacent lane, the controlled vehicle is deemed to be responsible.

[0049] The operation of the scenario parameter optimization unit 13 will be described with reference to Fig. 6. In step ST31, the scenario parameter optimization unit 13 corrects the objective function value D4 output from the objective function value calculation unit 11 based on the responsibility determination result D8 output from the responsibility attribution determination unit 122. For example, in a scenario in which surrounding vehicles a, b, and c exist as surrounding agents, if it is determined that the controlled agent is not responsible for the occurrence of a predetermined event for a certain surrounding agent, for example, surrounding vehicle b, a penalty value is added to the objective function value fb corresponding to the surrounding vehicle b, making it difficult to generate a similar scenario parameter value in subsequent optimizations.

[0050] In step ST32, the scenario parameter optimization unit 13 updates the values ​​of the scenario parameters by performing optimization. As an optimization method, there are heuristic methods such as Hill Climbing, Simulated Annealing, and Genetic Algorithms.

[0051] In step ST33, the scenario parameter optimization unit 13 outputs the updated scenario parameter values ​​to the simulation execution device 20.

[0052] As described above, the scenario parameter optimization device 10 assumes a scenario in which a plurality of agents including a controlled agent and surrounding agents exist. Under the assumed scenario, the scenario parameter optimization device 10 optimizes the scenario parameters such that the controlled agent is responsible for the occurrence of a predetermined event during a period of simulation testing of a control logic program, and such that the objective function value is smaller. Therefore, the scenario parameter optimization device 10 can effectively search for a scenario to be used as a test case in the simulation testing.

[0053] It is possible to combine the embodiments, and to modify or omit each embodiment as appropriate. [Industrial Applicability]

[0054] The scenario parameter optimization device disclosed herein can be used as a device for effectively searching for scenarios used to inspect a control logic program that controls an agent such as an autonomous vehicle. [Explanation of symbols]

[0055] 1 control logic inspection system, 10 scenario parameter optimization device, 11 objective function value calculation unit, 12 responsibility determination unit, 13 scenario parameter optimization unit, 20 simulation execution device, 100a processing circuit, 100b processor, 100c memory, 121 responsibility determination time detection unit, 122 responsibility attribution determination unit.

Claims

1. an objective function value calculation unit that receives an execution result of a simulation test of a scenario including a first agent and a second agent, which is executed based on a scenario parameter, and calculates an objective function value of a predetermined event from an objective function that expresses the predetermined event and an event occurrence time at which the predetermined event occurs, based on the execution result; a responsibility determination unit that determines whether or not the first agent is responsible for the occurrence of the predetermined event based on the execution result and the calculated event occurrence time from a situation at a responsibility determination time that is a time prior to the calculated event occurrence time, and outputs a result of the determination as a responsibility determination result; a scenario parameter optimization unit that optimizes the scenario parameters based on the calculated objective function value and the output responsibility determination result, so that the calculated objective function value becomes smaller and it is determined that the first agent is responsible for the occurrence of the predetermined event; A scenario parameter optimization device comprising:

2. the responsibility determination unit includes a responsibility determination time detection unit that detects the responsibility determination time based on the execution result and the calculated event occurrence time.

2. A scenario parameter optimization device according to claim 1.

3. the responsibility determination time detection unit calculates, as a candidate responsibility determination time, a time that is the latest before the event occurrence time among times at which a behavior change of the second agent occurs, and detects the calculated candidate responsibility determination time as the responsibility determination time.

3. A scenario parameter optimization device according to claim 2.

4. The change in the behavior of the second agent is either a change in course, acceleration, or deceleration.

4. A scenario parameter optimization device according to claim 3.

5. the responsibility determination time detection unit, when a recognition start time at which the first agent begins to recognize the second agent is later than the responsibility determination time candidate, detects the recognition start time as the responsibility determination time instead of the responsibility determination time candidate; 4. A scenario parameter optimization device according to claim 3.

6. the responsibility determination unit includes a responsibility attribution determination unit that determines whether or not the first agent is responsible for the occurrence of the predetermined event based on a situation at the responsibility determination time; the responsibility attribution determination unit determines whether or not the first agent is responsible based on a deceleration or a turning angular velocity of the first agent required to avoid the occurrence of the predetermined event at the responsibility determination time; 2. A scenario parameter optimization device according to claim 1.

7. the responsibility attribution determination unit determines that the first agent is responsible if a behavior change of the first agent occurs after the responsibility determination time; 7. A scenario parameter optimization device according to claim 6.

8. A scenario parameter optimization device according to any one of claims 1 to 7; a simulation execution device that performs a simulation test of the first agent using a scenario defined by the optimized scenario parameters and a control logic program that controls the behavior of the first agent; A control logic inspection system comprising:

9. A scenario parameter optimization method performed by a scenario parameter optimization device including an objective function value calculation unit, a responsibility determination unit, and a scenario parameter optimization unit, comprising: a step in which the objective function value calculation unit receives an execution result of a simulation test of a scenario including a first agent and a second agent, which is executed based on a scenario parameter, and calculates an objective function value of a predetermined event from an objective function expressing the predetermined event and an event occurrence time at which the predetermined event occurs, based on the execution result; the responsibility determination unit determining, based on the execution result and the calculated event occurrence time, from a situation at a responsibility determination time that is a time prior to the calculated event occurrence time, whether or not the first agent is responsible for the occurrence of the predetermined event, and outputting a result of the determination as a responsibility determination result; the scenario parameter optimization unit optimizing the scenario parameters based on the calculated objective function value and the outputted responsibility determination result, so that the calculated objective function value becomes smaller and it is determined that the first agent is responsible for the occurrence of the predetermined event; A scenario parameter optimization method comprising:

10. a function of receiving an execution result of a simulation test of a scenario including a first agent and a second agent, which is executed based on a scenario parameter, and calculating, based on the execution result, an objective function value of a predetermined event from an objective function expressing the predetermined event, and an event occurrence time at which the predetermined event occurs; a function of judging whether or not the first agent is responsible for the occurrence of the predetermined event based on the execution result and the calculated event occurrence time from a situation at a responsibility judgment time that is a time prior to the calculated event occurrence time, and outputting the result of the judgment as a responsibility judgment result; a function of optimizing the scenario parameters based on the calculated objective function value and the outputted responsibility determination result, so that the calculated objective function value becomes smaller and it is determined that the first agent is responsible for the occurrence of the predetermined event; and A scenario parameter optimization program that runs the above on a computer.

Images