Online generation of stable gait for biped robots with feedback loop algorithm

Abstract

The paper is devoted to the problem of stable gait generation for bipeds. The presented algorithm enables an online generation of stable gait patterns with minimal computational effort. The gait patterns are specified by the user with few task parameters: step size, step duration and walking speed, which can be changed in the generation process or during movement of the robot. The algorithm is based on a feedback loop scheme, composed of a linearized and decoupled non-linear model of the biped and time synchronized controllers. The closed form solution of the gait generation problem for a simplified model was considered to derive formulas for parameters required by the algorithm. Balancing/stabilization of the robot is achieved by appropriate change of the joint accelerations. The algorithm is applied to the 3D model of the biped. The performance of the algorithm was studied and demonstrated in a simulation. The algorithm can be used as a gait pattern generator, as a high level controller, and for balancing/stabilization of existing movement trajectories.