Adaptive reference management and model predictive control for near-surface depth–heading control of autonomous underwater vehicles

Near-surface operations conducted by underwater vehicles (UVs) feature strong nonlinear disturbances and speed-dependent control effectiveness, which make depth and heading regulation challenging under strict state and actuator limits. This paper presents an optimal–adaptive strategy on the Joubert BB2 platform that couples a constraint-aware model predictive controller (MPC) with an L 1 adaptive reference management architecture to mitigate unmodeled dynamics and uncertainties. The MPC is built from a reduced-order BB2 model that is linearized about forward speed into vertical and horizontal subsystems, and the infinite-horizon optimal control problem (OCP) is transcribed using Bernstein polynomial properties. Simulation results for two representative forward-speed cases show that the proposed L 1 -assisted MPC reduces the steady-state depth and heading errors observed with MPC-only control while respecting the prescribed actuator magnitude and rate limits in the tested scenarios.