Tracking Control for Streaming Input Motion Using Segmented Foot Model

Abstract

Motion capture data has been widely used to make realistic animation. Especially, in physics-based character simulation, motion data provides plausible reference trajectory of the character. However, it requires post-processing of the unnatural motion data such as foot-skating and use specified controller with different parameters for each motion. Recent advancements in motion capture hardware such as Kinect allow us to obtain various motion data low-cost and easy way in real-time, but still it remains challenge to simulate unpredictable motion with common controller. The human foot is complex structure containing 26 bones and it allows complex movements with balance. However, widely used human dynamics model has simple foot model which consist of one or two body, so it has functional limitations such as keeping flatfoot while static balancing. In our work, we propose new foot model that mimics the human foot to control various motion in realtime. Because foot contact condition with the ground is critical to maintain balance, more contact points and additional DOFs from segmented foot provide ability to control foot and whole body to be desired state in various foot contact condition.