Robot navigation by fusing a spatio-temporal video descriptor with a robust humanoid motion control for kicking a ball

Abstract

Cooperation work among robots is a very popular and active research area, in which the synergy among different dynamic methods, communication strategies and feedback control algorithms allows that a group of robots work together to achieve a common goal. Although, the last years have been reported important advances in this line, there exist many limitations on the develop of these collaborative works because the extremely variability of the scenes, the task challenge and some robot’s hardware issues. This work develops the specific tasks of locating, reaching and kicking a ball for a humanoid robot, and proposes a new approach which solves stability and illumination problems. Vision tracking feedback to locomotion algorithm will be employed in those tasks. Vision algorithm is based on spatiotemporal characterization, which give the input to the motion planning robot algorithm. The motion planning algorithm introduces a novel control approach which allows to reach the ball and kick it stably. A validation of the proposed method in terms of tracking algorithm error with motion control, in both real and simulated indoor environment allows to evidence the proper performance of the proposed approach.