Linear quadratic tracking based on reinforcement learning and motor speed control without system dynamics