Walking in the resonance with the COMAN robot with trajectories based on human kinematic motion primitives (kMPs)

摘要

Research in humanoid robotics aims to develop autonomous systems that are able to assist humans in the performance of everyday tasks. Part of the robotics community claims that the best solution to guarantee the maximum adaptability of robots to the majority of human tasks is mimicry. Based on this premise both the structure of the human body and human behavior have been the focus of studies, with the aim of imitating and reproducing on robotic systems the results of millennia of human evolution. The research presented in this paper aims (i) at transferring the features of human locomotion to the Compliant huMANoid (COMAN) robot, by means of kinematic motion primitives (kMPs) extracted from human subjects, and (ii) at improving the energetic performance of the walk of COMAN by exploiting its intrinsic compliance: it will be shown that, when the robot is walking at a gait frequency that is close to one of the main resonance frequencies of the mechanism, the springs contribute to tracking the human-like kMPs-based trajectories imposed, providing at the right time about 15% of the energy required for locomotion, and that was previously stored.