A Simultaneous Learning and Control Scheme for Redundant Manipulators With Physical Constraints on Decision Variable and Its Derivative

In this article, a simultaneous learning and control scheme built on the joint velocity level with physical constraints on the decision variable and its derivative, i.e., joint angle, joint velocity, and joint acceleration constraints, is proposed for the redundant manipulator control. The scheme works when the structure parameters involved in the forward kinematics are unknown or implicit. The learning and control parts are incorporated simultaneously in the scheme, which is finally formulated as a quadratic programming problem solved by a devised recurrent neural network (RNN). The convergences of learning and control abilities of the RNN are proved theoretically. Simulations and physical experiments on a 7-degrees of freedom (DOFs) redundant manipulator show that, aided with the proposed scheme and the related RNN solver, a redundant manipulator with unknown structure parameters can perform a given inverse kinematics task with high accuracy while satisfying physical constraints on the decision variable and its derivative.