Modeling and control of swimming gaits for fish-like robots using coupled nonlinear oscillators

Abstract

Fish have perfect body mechanisms and swimming modes for underwater locomotion, and provide ideal models for the propulsor design of fish robots. This thesis investigates modeling and control of swimming gaits for multiple degree of freedom (DOF) fish-like robots. An experimental-based approach is conducted, i.e. online generation and tuning of swimming gaits for fish robots according to its locomotion performance. The online modeling aims to achieve fish-like swimming locomotion on robots, and optimal swimming gaits associated to fast swimming speed and high energy efficiency of locomotion. The investigation consists of two parts. Firstly, the online gait generators are formulated by using artificial central pattern generators (CPGs). Secondly, the gait is tuned by using a closed-loop swimming control system incorporated with the CPG model. In order to obtain diversified gait patterns and gait transitions, different structure and coupling ways of CPGs are discussed so that feedback control laws can be imposed on the CPGs to shape the swimming pattern. The parameters of the oscillators are tuned as fish robots swims, without causing jerks or losing coordination of motions. The implementation of the system includes the combination of CPGs and closed-loop swimming control, which is initially applied in a special case of adaptive swimming: the station-holding control in adverse unsteady flow. Subsequently, the system is utilized, together with genetic algorithm (GA), to produce fast or energy-efficient swimming gaits for multi-DOF fish robots. Three types of fish robots have been deployed as experimental platforms for the research work here. Experiments are conducted to assess the CPG-based gait generation and the closed-loop swimming control system, including sine function based gait generators, robust gait generation and station-holding control of fish robots, multi-DOF swimming gait coordination and gait transition by using artificial CPGs, robust swimming gait tracking control, and online optimization of swimming performance by using GA and CPGs.