Feedback error learning, ffw ann compensates torque, feedback from pd to ann