Design of a Bio-Inspired Legged Robot for Rough Terrain Locomotion

Abstract

This research presents a unique biologically motivated control technique for synchronizing the 
planning of a quadruped robot’s gaits and foot Workspace Trajectory (WT), based on Central Pattern Generator 
(CPG) Neural Network (NN) WT (CPGNNWT). Initially, a foot workspace trajectory is designed using the 
Denavit-Hartenberg (D-H) nomenclature and inverse kinematics, offering benefits such as little mechanical 
shock, fluid motion, and streamlined trajectory. This work proposes an enhanced CPG using Hopf oscillators 
for seamless gait planning. A NN is ultimately developed and trained to transform the CPG output into the 
predetermined WT so that it can concurrently use the benefits of the CPG-based approach in gait prediction 
and the WT-based approach in foot trajectory preparation. Virtual prototype simulations and tests with actual 
quadruped robots are conducted to evaluate the efficacy of the suggested control technique. The findings 
indicate that the CPG effectively and efficiently regulates the quadruped robot's gait via internal variables. At 
the same time, the foot path aligns closely with the predetermined WT.