Bionic Control of a Roller-Skating Robot with Central Pattern Generator and Fuzzy-Q Learning

摘要

A bionic adaptive control method for a rollerskating robot is presented. The control system consisted of a high level Fuzzy-Q learning controller and a low level central pattern generator(CPG) neural networks. Each CPG is modeled as a modified Hopf nonlinear oscillator with direct mapping from parameters to outputs. By changing the parameters of CPGs, skating pattern, speed and direction can be changed correspondingly. The high level controller plays the role of changing the parameters of CPGs according to the Invironment condition. Fuzzy interface system is used to map the environment to CPG parameters and the fuzzy rule base is established by Q-learning. Navigation is elected to test the bionic system and the feasibility and effectiveness of the proposed method is confirmed by simulation results.