COLLISION CHECKING STRATEGY IN BIPED HUMANOID ROBOT NAVIGATION TOWARD OPERATION IN UNSEEN ENVIRONMENT

摘要

This paper presents new collision checking strategy in humanoid robot navigation. We have developed a contact interaction-based navigation system for a biped humanoid robot to improve current visual-based navigation and help humanoid robot operation in unseen environments. This system is based on contact interaction with objects using humanoid arms which equipped with six-axis force sensors. To ensure safety of humanoid robot in the proposed navigation tasks, we introduced collision checking strategy using arm contact approach considering biped walking characteristics in various directions. The collision checking is defined by searching motions of the robot's arms which created radius of detection area within the arm's reach. The detection areas are analyzed using Rapid-2D software to define absolute collision free area which used to calculate maximum step-length in biped walking. In case of object is detected during searching motion, the robot arm will touch and grasp the object surface to define self-localization, and consequently optimum step-length is refined. Verification experiments using humanoid robot Bonten-Maru II to operate in a room with wall and obstacles was conducted, which results revealed good performance of the robot to realize collision free area and generate optimum biped walking.