Event-Driven Fault Injection Testing Framework

As a highly complex embedded system, an autonomous driving system integrates key technologies such as sensors, perception, planning, and control to accomplish crucial tasks like real-time environmental perception, path planning, and vehicle control. Due to the complexity and variability of environmental information, the perception module is prone to generating fault information and transmitting it to the planning control module. This can lead to erroneous decisions and control commands by the planning control module, ultimately causing accidents. Therefore, it is crucial to test the tolerance of the planning control module to fault information. This paper proposes an event-driven fault injection testing framework that can dynamically inject various types of faults based on different fault trigger conditions in the scenario. This approach effectively evaluates the performance of autonomous driving systems under diverse fault conditions. The experimental results show that the framework successfully achieves dynamic fault injection and can clearly define the fault tolerance boundaries of the planning control module.