Model-Predictive Path-Integral Control of an Unmanned Surface Vessel with Wave Disturbance

High-agility maneuvers within the maritime domain require situational awareness of sea state and other environmental disturbances. Such disturbances are generally not measurable directly, instead informed through a high degree of uncertainty and inference through the reactivity of the surface vessel. In this work, we implement a model-predictive path-integral control (MPPI) framework to provide highly agile control of the Clearpath Robotics Heron unmanned surface vessel (USV). In addition to increasing scope of MPPI control to allowing external wave forces, the implementation improves upon the state-of-the-art methods of MPPI control for such maritime systems as well. We discuss the uncertain dynamics both within the USV and the maritime environmental forces, derive an MPPI-based control framework, and finally discuss learning-based methods for tuning the controllers with limited information. Simulation results are provided to demonstrate the efficacy of the system.