Adaptive speed control for autonomous surface vessels

The paper addresses the problem of speed control for the SeaFox unmanned surface vessel (USV). This small, versatile robotic platform can operate over a wide range of speeds, making it attractive for a number of scientific, commercial, and naval applications. This versatility, however, comes at a price. The vessel operates in displacement mode at low speeds and operates in planing mode at high speeds. These two regimes are connected via a highly unstable transition mode, where steady state operation is not possible, making autonomous operations challenging. Speed following is one of the key challenges in automating this class of vessel, as this capability is adversely affected by (i) the inherently slow dynamic response of the propulsion system, (ii) significant variation of the vessel's hydrodynamics in three distinct operating modes, and (iii) significant coupling between these hydrodynamics and the propulsion force. This paper presents a comparative study of three adaptive control algorithms developed for speed-holding capability on the SeaFox USV: (i) classical PID control with gain scheduling, (ii) model reference adaptive control, and (iii) L1 adaptive control. Beginning with a description of the system identification experiments that informed our understanding of the open-loop plant dynamics, this paper proceeds through controller design and simulation, and presents results from open ocean sea trials. The experimental results provide a basis for an objective comparison of each algorithm's speed following performance and explicitly highlight the benefits of adaptive controllers.