自动驾驶汽车感知系统测试用例构建方法研究

自动驾驶汽车对感知系统、决策系统和执行系统的功能、性能和可靠性提出了更高的要求。这些要求都需要通过试验验证来进行确认,并根据试验结果对开发过程进行指导和改进。在现实世界中,高度复杂的驾驶环境使得试验的测试用例数量变得异常庞大,因此如何设计合适的试验方法,用尽可能少的测试用例获得更多的测试信息,是一个非常重要的研究课题。本文分析了预期功能安全7层场景库架构对于感知系统的影响,根据其对感知系统的影响程度选取了适用于不同系统的试验水平和因子,并基于感知系统的测试评价指标,设计了正交设计试验方案,该方案将在保证测试目的的前提下提高测试效率。

Autonomous driving vehicles set higher requirements on the function,performance,and reliability of perception systems,decision-making systems,and execution systems.These requirements need to be confirmed through experimental verification,and the development process can be guided and improved according to the experimental results.In the real world,the highly complex driving environment makes the number of test cases extremely large.Therefore,how to design a suitable test method and obtain more test information with as few test cases as possible is a very important research topic.The paper analyzes the impact of the expected functional safety 7-layer scene library architecture on the perception system,selects the test levels and factors suitable for different systems according to the degree of its impact on the perception system,and designs an orthogonal test based on the evaluation index of the perception system.Such test plan will greatly improve the test efficiency with ensuring the test purpose.