Locomotion Control of Robotic Fish with a Hierarchical Framework Combining Spiking Neural Networks and CPGs

Abstract

Controlling a central pattern generator (CPG) network to generate multi-mode signals is a key issue for a self-propelled bionic robotic fish. In this paper, a robotic fish oriented hierarchical control framework combining CPGs with spiking neural networks (SNN) is proposed. Specifically, as a movement mode generator, the CPG network comprises a weak coupling chain of phase oscillators. The output signals of the CPG network serve to regulate locomotion modes of joints of robotic fish. Meanwhile, the CPG network receives signals from the SNN. In particular, the SNN serves as a higher controller, receiving sensor signals and output mode commands to the lower generator. The entire control procedure is similar to one of human movements. Simulation results on diversified swimming modes verify the feasibility of the hierarchical control framework integrating the CPGs with SNN.