Terrain-Adaptive Model Predictive Control for Autonomous Off-Road Vehicles in Unstructured Environments

Autonomous navigation in unstructured off-road environments faces fundamental challenges from terrain-induced dynamics, multi-axle complexity, and computational intractability. This paper proposes a Terrain-Adaptive Model Predictive Control (TA-MPC) framework, which constructs a 3D Frenetbased parametric surface to encode real-time elevation and employs a semi-explicit differential-algebraic equation (DAE) model to specifically resolve high-order statically indeterminate load transfer issues and dynamic modeling issues of multi-axle vehicles under nonplanar constraints through suspension dynamics and wheel decoupling. This framework bridges a critical gap in terrain-adaptive control algorithms for multi-axle heavy-duty vehicles. Validated via TruckSim-MATLAB co-simulation on an 8 × 8 all-wheel-drive heavy-duty vehicle, TA-MPC demonstrates remarkable performance under extreme operating conditions.