The Influence of Learning Time on CPGs with Hopf Oscillators inBiped Locomotion

In this paper we study the effect of learning time innonlinear dynamical control systems designed for locomotionmovement systems, more specifically biped locomotion system.Due to the complexity of rhythms in locomotion movements,Central Pattern Generators (CPGs), inspired from nature, havebeen studied to provide a robust mechanism for producing therequired rhythms. Artificial CPGs are composed of adaptiveHopf oscillators coupled together based on a neural networkstructure. Every Hopf oscillator has equipped with learningsection, and through learning from its input signal, it is able toadjust its oscillation frequency and phase according to thefrequency and phase of its input signal. Our aim here is to studythe learning time and its trade-offs in accordance withpreserving the stability and trajectory of the biped locomotionsystem against perturbations.

Keywords:

Adaptive Control , Biped Walking , CPG , Hopf Oscillator , Humanoid Robot , Trainable Oscillator