Humanoid robot walking on track using computer vision

Abstract

This paper is addressed to a problem of humanoid robot locomotion with the use of computer vision. The robot is 20 DoF Darwin OP equipped with video camera in its head and has built-in PC. To determine the desired path we use the algorithm of video processing based on image filtration within threshold values in HSV color model. Developed control system includes tracking system for the head with camera and speed PD-controllers for robot motion. To verify system performance we use a side video camera mounted under track and process its data in the similar way. Filtered data from the side camera allows to measure position error between the robot and desired path. Experiments are held for three different trajectories (curve in the form of butterfly, circle and square) and show the efficiency of the proposed solution.