Abstract: Kinematical optimization of redundant robots is proposed to be simultaneous with dynamical optimization according to the characteristics of their redundancies. The joint torques of a redundant robot are rationally distributed dynamically by the weighted least norm algorithm of acceleration, while the avoidability of singular positions of a redundant robot is optimized kinematically by gradient projection algorithm, and an algorithm of simultaneous solution is proposed to both kinematical and dynamical optimization. Simulation results showed that the algorithm is able to not only distribute rationally the joint torques but also optimize kinematically the motion of robots, thus implementing the simultaneous optimization of the redundant robots kinematically and dynamically.