Human–machine collaborative driving collision avoidance control based on Bayesian risk estimation

How to address the challenges in human–machine cooperative control of intelligent vehicle is still an important issue, which include the conflicting objectives between driver and controller, the difficulty in fusing multi-source risk information from both vehicle and driver states, the limited robustness of conventional risk estimation under uncertain interference. To this end, this paper proposes a human–machine collaborative driving collision avoidance control based on Bayesian risk estimation. Firstly, a Bayesian multi-factor driving risk estimation model is designed by integrating time-to-collision, collision distance and driver fatigue state. Simultaneously, the risk level is used as controller weights within the optimization objective of the non-cooperative game-theoretic controller for resolving conflicting objectives between the driver and the controller. Finally, the effectiveness of the proposed method is verified using PreScan and Simulink co-simulation, along with a hardware-in-the-loop driver testing platform. Experimental results show that under single and double-lane-change scenarios, the proposed strategy exhibits strong disturbance resistance in the white noise and fast response, significantly improving collision avoidance performance in complex environments.