Rapid Generation of Challenging Simulation Scenarios for Autonomous Vehicles Based on Adversarial Test

As the development of the autonomous vehicles, the test and evaluation of the autonomy of autonomous vehicles has receiving considerable attentions from both academic and industrial research fields recently. How to rapid identify the performance boundaries of autonomous vehicles and then to accelerate the testing has been considered as one of the most important problems. Simulation test and challenging scenarios are considered as two effective ways to solve above problems. In this research, a framework of rapid generation of challenging simulation scenarios based on adversarial test was proposed. An optimization model based on newly developed indicators, i.e., safety, smoothness, sharpness and smartness, was proposed to assess the autonomy of autonomous vehicles. Due to the high dimensionality of the configuration space and computation complexity of the optimization problem, a quantum genetic algorithm was employed to solve the optimization problem owing to its high search ability and lower population size required. A simulation system based on Prescan, Carsim and Simulink was built to validate the effectiveness of the proposed method. Comparative experiments also were conducted to illustrate the efficiency of the proposed optimization algorithm.