Wave-filtering finite-time self-learning extended state observers for robotic surface vehicles

Abstract

This paper presents wave-filtering finite-time self-learning extended state observers for robotic surface vehicles subject to unknown control gains and ocean disturbances. By incorporating wave filtering technique and concurrent learning into the extended state observer design, the low-frequency motion components and wave-frequency motion components of robotic surface vehicles can be estimated, in addition to the unknown control gains, total disturbance and unmeasured velocity. A salient feature of the proposed approach is that wave-frequency motion components are filtered from the position and yaw angle sensor measurements to avoid mechanical wear and tear of the propulsion system components caused by wave-frequency motion signals. The result is extended to a wave-filtering finite-time self-learning extended state observer which improves the estimation performance considerably. The efficacy of the proposed two wave-filtering self-learning extend state observers for robotic surface vehicles with unknown control gains is substantiated via simulations.