Control of humanoid robot walking by Fuzzy Sarsa Learning

This paper proposes applying Fuzzy Sarsa Learning (FSL) algorithm to biped walking controller of a humanoid robot for having a stable walking motion. In order to evaluate stability of the biped walking in the control problem, we have used the Zero Moment Point (ZMP) criterion based on the inverted pendulum model; via this modeling, the stability of biped robot will depends on trunk motion. The FSL controller has been applied in order to decrease ZMP error and consequently enhance the stability of biped walking through trunk motion. Humanoid robot has been defined as a seven-link biped model which is simulated as a set of equations of motion in MATLAB. Finally, the simulation results demonstrate the better stability and control of the walking motion by the FSL method compared with Fuzzy Q-learning (FQL) controller and basic Fuzzy controller.