Experimental evaluations of motion stabilization using image feature tracking for biped walking robot

Abstract

This paper presents an approach to stabilization control using image feature tracking for biped walking robot. The authors have developed several vision-based control for dynamic balance and stabilizing the body motion of biped walking robot. For example, the ground reaction force was estimated from visual target position in camera image. However, in our past developments, some landmark object is placed in order to evaluate the control performance with easy image processing. In this paper, KLT tracker (Kanade-Lucas-Tomasi Feature Tracker) is used for the ground reaction force estimation without any landmark object in visual scene, and the body motion stabilization is evaluated by using the feedback of estimated reaction force through several experimental results.