Keyframe Movement Optimization for Simulated Humanoid Robot Using a Parallel Optimization Framework

Abstract

Motion control for a high degrees of freedom humanoid robot is one the hardest problems in Mobile Robotics. Researchers have been very successful in designing walking motions using reduced order mathematical models based on the ZMP approach. However, this approach is unable to design very dynamic motions, such as movements to get up and long distance kicks. In these cases, keyframe movements have proven to be a successful technique. Therefore, this paper presents a system that allows a parallel approach for optimizing keyframe movements for the Robocup 3D Soccer Simulation League. Moreover, optimization results obtained using this system are shown to validate it.