State machine-based controller for walk-halt-walk transitions on a biped robot

Abstract

In this paper, we show how it is possible to obtain a walk - standup - walk cycle with a biped robot using only a state machine controller. This work is a continuation of another in which the same approach was used to control the walking gait of the same robot. We introduce four critical angles that affect robot speed and step length. Our control approach consists of two control levels. The low level is a PID controller and the high level is a dynamic walking algorithm, inspired from human locomotion, which tunes the PD controllers and impose the reference signals based on the walking which increases walk stability. This method could be easily implemented in real time because it needs acceptable calculation time. We validated the control approach to a dynamic simulation of our 14DOF biped called ROBIAN. We prove that we can maintain robot walking and stopping stability and walk cyclepsilas repetition without referencing a predefined trajectory or detecting the center of pressure. Results show that the walk of the biped is very similar to human one.