Motor Learning for Manoeuvring Control of a Remotely Operated Vehicle

The paper investigates the application of a bio-inspired control paradigm encompassing nonlinear function approximation and adaptive learning to design a manoeuvring controller for a remotely operated vehicle. The bio-inspired learning controller is adopted to overcome limitations introduced by unknown nonlinear dynamics as well as by environmental disturbances. The proposed controller is compared against a traditional adaptive controller with respect to tracking performance and control effort. Simulation results show that the bio-inspired controller displays better tracking performance when the structure of the system dynamics is unknown. Further, the paper shows that offline training of the functional approximator helps in delivering a smoother control effort, and that a bio-inspired integral action fulfils the disturbance rejection.