Studying the Effect of Different Optimization Criteria on Humanoid Walking Motions

摘要

The generation of stable, efficient and versatile walking motions for humanoid robots is still an open field of research. Several approaches have been implemented on humanoids in the past years, but so far none has led to a walking performance that is anywhere close to humans. This may be caused by limitations of the robotic hardware, but we claim that it is also due to the methods chosen for motion generation which do not fully exploit the capabilities of the hardware. Often, several characteristics of the gait, such as foot placement or step time, are fixed in advance in a suboptimal way for the robot. In this paper we discuss the potential of our optimal control techniques based on dynamical models of the humanoid robot for the generation of improved walking motions. We apply the method to a 3D dynamic model of the humanoid robot HRP-2 with 36 DOF and 30 actuators. Robot specific stability constraints (such as ZMP constraints) can be taken into account in the optimization. We present results for five different objective functions, and evaluate the influence of free foot placement and a relaxation of ZMP constraints.