Vision Guided Motion Control of a Biomimetic Quadruped Robot: RoboCat

This paper presents the vision system and visual processing for a biomimetic elastic cable-driven quadruped robot—RoboCat. The paper is geared towards selection of appropriate visual servoing techniques for RoboCat such as vision algorithms, high-level cognition algorithms, software architecture and hardware implementation. The system uses two video cameras for stereo vision data acquisition and a SUMIT-ism form factor embedded computer for vision data processing. The vision system employs a color based target recognition algorithm, a neural network based shape recognition algorithm and a Color and Zernike moment based face detection algorithm. The paper presents the vision algorithms, vision guidance and motion tracking algorithms, rule-based decision making algorithms and the open architecture of the autonomous vision tracking system. Experimental testing results (including video clips) are also presented.