A Reliability Test Scenario Generation Method for Unmanned Vehicle Off-Road Tracking Task Based on Deep Reinforcement Learning

It is important to conduct simulation tests of unmanned vehicles (UV). However, some test scenario generation methods face challenges such as a lack of off-road scenarios and insufficient consideration of interferences. This paper proposes an unmanned vehicle off-road task reliability test scenario generation method. First, the static test environment is built based on the TORCS simulation platform. Subsequently, different types of unmanned vehicle-target vehicle (TV) tracking interaction behaviour models are generated based on the modified DDPG method. Finally, the unmanned vehicle task reliability test under different disturbances is achieved. The results show that the method can not only effectively generate off-road tracking test scenarios, but also adjust the interference to test unmanned vehicles' task reliability. The method provides valuable guidance for testing and enhancing the reliability of the model to some degree.