Biped walking pattern generation based on spatially quantized dynamics

Abstract

We present a biped walking pattern generation based on a new way of system discretization named spatially quantized dynamics (SQD). In SQD, a continuous system is discretized by a constant unit length along the walk direction, and the dynamics is represented by a recurrence formula for a unit length motion, taking variable period for each cycle. Using SQD modelling, we can generate a biped gait by taking three steps;1)Design a walking pattern only considering kinematics in the sagittal plane, 2) Calculate the ZMP and velocity by optimization of the spatially quantized dynamics, 3) Add hip lateral motion for 3D dynamic balance. It is shown that we can easily generate a biped gait with stretched knees and a gait for large step climbing. The validity of the generated patterns are confirmed by simulations.