Data-Assisted Modeling and Speed Control of a Robotic Fish

Abstract

In this paper, a novel data-assisted dynamical modeling and control approach is developed for robotic fish speed tracking. The data-assisted modeling focuses on the thrust mechanism, including the structure and parameters that are absent from the Newtonian-based analytic model of the robotic motion. The thrust of a robotic fish is generated through undulatory body movement interacting with surrounding water, thus a consequence of reaction from environmental hydrodynamics. It is known, however, that hydrodynamics cannot be analytically modeled. Thus, the data-assisted modeling is necessary for an underwater robotic fish. Specifically in this work, data of pulse and step responses are collected from designated experimental trials, in which the pulse responses are used to determine the thrust delay terms, and step responses are used to build up the thrust nonlinearity at steady state. A discrete-time sliding mode controller (SMC) is constructed to perform speed control. The experimental results verify that an SMC with a data-assisted model can substantially improve the speed control performance of two-dimensional robotic motion.