A Novel Design Methodology of CPG Model for a Salamander-Like Robot

Abstract

In this letter, we present a novel method for the design of central pattern generator (CPG) model, and then utilize it to control a salamander-like robot. The CPG network is composed of modified Hopf oscillators with a new coupling scheme that independently controls the waveform regulation and phase coordination process. To improve the accuracy of waveform regulation, corresponding coupling weight parameters are dynamically adjusted using an adaptive learning rule, and ultimately reach their critical value. Besides, the CPG network can achieve fast convergence by employing this critical value with an updating rule. Afterward, the designed model is utilized to develop a control strategy for the salamander-like robot, involving gait characterization and foot trajectory generation. The control strategy is capable of producing several terrestrial gaits with smooth transitions among them. Finally, numerical simulation and hardware experiments are performed to demonstrate the validity of the proposed method. Moreover, this method can be applied to a variety of biologically inspired robots, where asymmetric oscillation signals with phase coordination are necessary.