Recursive Inverse Kinematics Analysis for Teaching Human Motion to a Humanoid Social Robot using a Depth Camera

Teaching human motion to a humanoid social robot is important because it enables a humanoid social robot to interact with humans more naturally and in a friendly manner. Recently, a depth camera-based human motion teaching has been developed. However, inverse kinematic analysis using human motion skeletal data obtained from the depth camera is not general but rather robot structure-dependent. Thus, in this paper, we present a more general approach to compute joint angles from the skeletal data obtained from a depth camera using a recursive inverse kinematic algorithm. With a known base body orientation, each joint angle is recursively computed from the base body (torso) to the tree end body (robot hand). To validate the proposed recursive inverse kinematic algorithm, simulations of a humanoid robot model have been carried out using the RecurDyn multibody analysis software. The virtual humanoid robot consists of two 6 DOF arms and a 2DOF waist. The human upper body motion when walking has been captured using a Microsoft Azure Kinect depth camera. The hand motion from the depth camera and that from the simulation have been compared to investigate the effectiveness of the proposed algorithm.