[2P4-GS-11] Robot and real worlds: Social application

[2P4-GS-11-02] Automatic generation of dynamic behavior data of surrounding vehicles to ensure the reliability of self-driving vehicle

〇Yoshifumi Sakamoto¹, Kentaro Aota¹ (1. IBM Japan Ltd.)

Keywords:Self-Driving Vehicle, Simulation, Automatic generation, Test Cases

Simulation technology that guarantees reliability is widely used in the development of self-driving vehicle. The simulation often focuses on the function of Path Planning. The reason is that it is difficult to reproduce in the real world because the Path Planning depends on the state of the self-driving vehicle and the surrounding environment. What is important is the reproduction of the behavior of the vehicles around the self-driving vehicle.

On the other hand, self-driving vehicles are required to operate properly in situations such as accidents. However, it is difficult to identify all situations in advance. It is not hard to imagine that a serious accident is likely to occur in cases that cannot be identified in the design. We have developed a co-simulator that automatically generates test cases that satisfy unintended test spaces and reproduces the dynamic interaction between self-driving vehicles and surrounding vehicles. In this co-simulator, the amount of calculation was reduced by adopting Active Matter, and as a result, the behavior of surrounding vehicles and the interaction with self-driving vehicle were automatically and real-time generated.