A neural computational scheme for infinity-norm joint torque minimization of redundant manipulators with actuator constraints

A neural network based on the projection and contraction method is applied for a redundant manipulator bounded minimum infinity-norm joint torque computation. The nonlinear joint torque optimization problem is transformed to a linear program which can be solved by the proposed neural computational scheme in real-time. While the desired accelerations of the end-effector for a given task are fed into the network, a driving joint torque vector which never exceeds the actuator limits and whose maximum component in magnitude is minimized, is generated as the neural network output. The proposed neural torque control scheme is shown to be capable of effectively generating the bounded minimum infinity-norm driving joint torques of redundant manipulators