Off-road distributed-drive electric vehicle trajectory tracking control with constrained Transformer MPC

Trajectory tracking control for off-road distributed-drive electric vehicles presents significant challenges due to compound slope effects and rollover risks. Existing approaches often neglect the critical impact of coupled slopes on vehicle roll dynamics and face substantial computational burdens during real-time implementation. This paper presents a four-degree-of-freedom vehicle dynamic model that comprehensively captures longitudinal, lateral, yaw, and roll motions while accounting for compound grade effects. We propose a novel Constrained Transformer Model Predictive Control algorithm that enables real-time policy computation while maintaining safety constraints. CarSim co-simulations demonstrate that our approach effectively prevents rollover and improves trajectory tracking accuracy by 72.28% across various off-road scenarios while reducing computational complexity by 213 times compared to conventional online optimization MPC. Real vehicle tests in off-road environments further validate the effectiveness of the proposed algorithm.