Path-following in USVs using LPV NMPC: Experimental validation in a propeller-driven boat

Path-following of unmanned surface vehicles (USVs) is often handled via Line-of-Sight (LOS) schemes. However, these strategies render suboptimal performance due to the inherent nonlinear and time-varying nature of the involved phenomena (wind speed variation, ocean currents, etc). Accordingly, we propose a dedicated Nonlinear Model Predictive Control (NMPC) algorithm for path-following in USVs. In particular, we identify an LTI Nomoto model, and then embed the lateral motion nonlinearities using a quasi-Linear Parameter Varying (qLPV) approach. Our scheme is experimentally validated in a 32-foot propeller-driven sailboat ( Brasilia 32 ), using the LPV-PILOT NMPC software. The boat is instrumented using low-cost equipment (a DC motor coupled to the helm wheel, GPS measurements for position, a potentiometer for rudder angle measurements and an ESP32 board for embedded control). As a side contribution, we also tune and validate a virtual sensor that estimates the USV’s rudder angle. Our experimental results indicate the effectiveness of the scheme in a low-cost setup.