Recursive inverse kinematic analysis for humanoid robot based on depth camera data

We propose the recursive inverse kinematics (RIK) algorithm to compute joint angles of a humanoid robot using skeleton tracking data. Recursive inverse kinematics calculates the angles recursively from the base body (torso) to the tree-end body (hand or foot or head) along each branch of the humanoid robot based on the multibody analysis approach. To validate the RIK algorithm, experiments with two specific human motions were conducted in an indoor environment with the Azure Kinect depth camera. The joint angles obtained from RIK with the skeleton data of human motion were compared to those from RIK with the corresponding virtual skeleton data of the RecurDyn virtual humanoid robot model. Additionally, similarity indices were defined to evaluate how similar the movements of the virtual humanoid robot from the RIK are to those of human. The obtained similarity indices showed that the posture of the virtual humanoid robot model using the RIK is close to the human posture.