An efficient multi-model predictive control for path-following of unmanned surface vehicles across wide speed ranges

Wide-speed-range maneuvering aggravates the nonlinearity in the dynamics of unmanned surface vehicles (USVs), making single-model-based controllers difficult to guarantee performance. This paper presents an efficient multi-model predictive path-following control strategy. The wide-speed-range nonlinear dynamics of the USV are modeled as a speed-scheduled piecewise affine system. Using time-scale separation, the speed-heading model is decoupled, with model errors and disturbances treated as lumped disturbances. We develop an improved adaptive heading guidance strategy to generate the desired heading, embedding speed into adaptive laws to boost USV dynamic adaptability across wide speed ranges. A combined feedforward-feedback heading controller is proposed, using a finite-time lumped-disturbance observer for feedforward compensation. The feedback employs an efficient MPC with offline gain design and online gain scheduling, partitioning the heading state space precisely via quadrilateral invariant sets, which yields lower conservatism than traditional ellipsoidal sets. Simulations show superior performance, with heading control efficiency improved by over 96%.