Improved Online Kick Generation Method for Humanoid Soccer Robots

Approaching and kicking the ball toward the goal is the very essence of robotic soccer. However, for humanoid robots, it is hard to make a fast and stable kick as it includes large momentum changes while standing on a single foot. Simple keyframe based approaches tend to fail due to differences between robots or external perturbations, so they need to be executed slowly. Slow execution increases the risk of losing the ball preparing for the kick, however. In this paper, we first review two online kick generation methods that we have been using for the RoboCup competition: a stationary kick controller that uses inverse kinematics and feedback stabilization and a walk-kick controller which is based on our analytic zero moment position (ZMP) based walk controller. Finally, we present a ZMP preview based kick generation method that combines the power of the stationary kick with the speed of the walk-kick controller.