自动驾驶汽车仿真测试与评价方法进展

Progress in Virtual Test and Evaluation of Autonomous Vehicle

随着汽车自动驾驶系统的不断升级,如何高效地测试和验证自动驾驶算法的合理性及系统稳定性已经成为业内一大挑战。为了应对未来更高级别的自动驾驶汽车测试,作者评述并比较了五种主流的自动驾驶仿真测试方法的最新进展,包括蒙特卡洛方法、蒙特卡洛加速方法、博弈论方法、测试场景矩阵方法、驾驶模拟器仿真测试方法。系统分析后认为,利用驾驶模拟器结合其他测试方法可以满足未来更高级别自动驾驶车辆的测试需求。

With the upgrade of autonomous driving systems, it has become a major challenge in the industry to test and evaluate the control algorithm for safety and stability. In this review, to address the needs of future higher- level autonomous driving system, 5 virtual test methods are discussed and compared. These methods are the Monte Carlo method, accelerated Monte Carlo evaluation method, game theory method, test scenario matrix method, and driving simulator method. It is concluded that the use of driving simulator combined with other test methods can meet the requirements for the future test and validation of higher-level autonomous vehicles.