An Inverted Pendulum Model Based Humanoid Gait Generation Using Convex Optimization ?

Convex optimization has the advantages on solving the problems with hundreds of variables and thousands of constraints. However, traditional multi-body model based zero moment point (ZMP) formulation makes it impossible to introduce the convex optimization into humanoid gait generation because the constraint can not be transferred into convex. This paper derives a new ZMP formulation by combining the linear inverted pendulum model and the multi-body model in the representation of Lie group. With this new ZMP formulation, an instance of walking gait generation by using convex optimization is given. The experimental results show that the number of iterations involved in solving the problem is less than 10, which ensures the effectiveness and efficiency of using convex optimization to generate humanoid gait online.