Optimization based gait pattern generation for a biped robot

This paper addresses the problem of gait pattern generation for biped robots. A method based on nonlinear parameter optimization is proposed. Key features of the method are the trajectory description, optimization of only a subset of coordinates and the use of analytical, recursively calculated gradients for cost functions and constraints. The method has been used successfully to calculate walking trajectories for the 3D 17-DOF robot "Johnnie"