A gait adaptation scheme for biped walking robots

Recent years witnessed a growing interest in biped walking robots because of their advantageous use in the human environment. However, their control requires many problems to be solved because of the many degrees of freedom and nonlinearity in their dynamics. The so-called open loop walking with offline trajectory generation is one of the control approaches in the literature. There are various difficulties involved in this approach, the most important one being the difficulty in tuning the gait parameters. This paper proposes an online fuzzy adaptation scheme for one of the trajectory parameters in the offline generated walking pattern. A fuzzy identifier system, represented as a three-layer feed-forward neural network is employed to compute the parameter as a function of time in simulations. Fuzzy system parameters are adapted via back-propagation. Virtual torsional springs are attached to the trunk center of the biped. The torque generated by the springs serve as the criterion for the tuning and they help maintaining a stable and a longer walk which is necessary for the online tuning process. 3D simulation and animation techniques are employed for a 12-DOF biped robot to test the proposed adaptive method.