State estimation for autonomous surface vehicles based on Echo state networks

Abstract

This paper investigates the state estimation for autonomous surface vehicles in the presence of unknown dynamics and unmeasured states. The unknown dynamics comes from parametric model uncertainty, unmodelled hydrodynamics, and external disturbances caused by wind, waves and ocean currents. A nonlinear adaptive observer is proposed based on echo state networks, which are used to approximate the unknown dynamics using input-output data. By using the proposed observer, the unmeasured states and unknown dynamics can be simultaneously estimated in real time. The stability of the observer is analyzed via Lyapunov analysis. The proposed observer can be used in various motion control scenario, such as target tracking, trajectory tracking, path following, formation control, and even sideslip angle identification, not only for fully-actuated marine vehicles but also for under-actuated marine vehicles.