A hybrid inverse kinematics framework for redundant robot manipulators based on hierarchical clustering and distal teacher learning

Inverse kinematic models are among the most significant tools in robotics and a real time framework to solve the inverse kinematics is necessary for the robot to perform required task. However, for redundant robot arms with more degrees of freedom than required for a given task, the inverse kinematics remains a difficult and challenging problem. With the extra degrees of freedom, redundant robot arms can be much more flexible and dexterous than traditional non-redundant manipulators, thus are very suitable for performing many challenging tasks, such as grasping novel objects in flight and conducting human surgeries. In this work, a distal teacher learning framework combined with hierarchical clustering algorithm is proposed to solve the inverse kinematics of redundant robot arms. The hierarchical clustering algorithm is used to learn the inverse kinematic model of the robot, and the prediction error of the learned model is compensated by the distal teacher. Simulations in MATLAB performed on a five-degrees of freedom planar redundant robot have verified the effectiveness and efficiency of this method.