Human-Vehicle Cooperative Driving with Multi-Lane Decision-Making via Incomplete Information Game

Nowadays, human-vehicle cooperative driving is still an important solution for intelligent transportation in unstructured scenarios. However, the conflict between the driver and the automated system for multi-lane decision-making remains to be solved. When leaving the intended path, driver's expected trajectory cannot be identified by the automated system, re-sulting in safety hazards related to cooperation failure under incomplete decision information. This paper proposes a trajectory probability game control (TPGC) method to model the situation and intend to relieve the conflict. A Bayesian Nash equilibria strategy is derived using the proposed probabilistic dual iterative approximate dynamic programming (PADP). The equilibrium indicates a stable state for the human-vehicle system given the probability of driver's trajectory selection, where the uncertainty of the driver is reduced. A multi-lane simulation scenario is constructed to verify the effectiveness of the proposed control and solution method. The experimental results indicate that the designed controller can be applied to various trajectory selections and maintain an average lateral error within 0.17m on multi-lane roads.