Virtual-joint based motion similarity criteria for human–robot kinematics mapping

Abstract

Motion mapping is an important part in human–robot cooperation. In this paper, a novel concept of virtual-joint based similarity criteria is proposed for flexible and efficient kinematics mapping between dissimilar embodiments, including different degrees of freedom (DOFs), different body morphology, and so on. Virtual joints are defined respectively in both the demonstrator and the imitator, with the same number. In virtual joints, the neglecting, re-ordering and repetitive usage of DOFs could be realized through the virtual decomposing matrices. Each virtual joint of the demonstrator and the corresponding one of the imitator formed a virtual joint pair. The Total Metric of Motion Similarity is the weighted sum of the metrics defined for each virtual joint pairs. Unlike traditional joint-space or Cartesian-space based metrics describing motion similarity solely at the DOF kinematic mode level, virtual-joint-based metrics can be adopted to describe different aspects of motion similarity between dissimilar agents, both in joint space and in Cartesian space. Two experiments are conducted to illustrate the effectiveness of the proposed approach.