Abstract
We propose the planning method to directly plan the 3D gait of biped robot without path planning. The discrete-time kinematic model that represents the biped walking by pivoting and both legs interval expansion and contraction is employed as the geometric foothold selection model of bipedal. The gait planning problem can be treated as the inverse kinematic problem on this model, and we propose to solve the constrained nonlinear optimization problem instead of the IK problem. An evaluation function consists of the goal-state tracking square error, and the constraints are formulated in contacts of field, available foot place and angle on the basis of the humanoid robot NAO. We can obtain the optimized gait plan that strictly satisfies the constraints in real time by solving this optimization problem using iterative method that uses the Quadratic Programming method with constraints.
