Abstract: In order to solve the problem of control performance degradation of the robot due to the payload changes, an identification method by driving only the 3rd, 4th, 5th and 6th axes of the robot along the excitation trajectory is proposed, based on the analysis of the effect of payload dynamics parameters on the torque of each joint. Firstly, the robot dynamics model is linearized based on the minimum inertia parameter set. Secondly, the optimal excitation trajectory is designed with the finite Fourier series for payload identification based on the analysis of the effect of the payload parameters on each joint torque, after selecting corresponding joint axis. Then the data of joint angle and joint torque are collected respectively while the excitation trajectory is run in conditions of no payload and other three different payloads, before they are processed by the low-pass filter. Finally, the weighted least square method is used to identify the payload dynamics parameters based on dynamic linear model. After the robot runs the verification trajectory, the payload identification result is evaluated by calculating the root mean square (RMS) of the difference between the calculated payload torque and the measured payload torque. At the same time, this method is compared with CAD (computer aided design) method. Results show that the former method can reduce the RMS value of the latter up to 16%, and that it is stable and effective under different payloads. Using this method, the way to drive all the joint axes is avoided, and thus not only the error caused by the coupling of the robot, but also the optimization time of the excitation trajectory are reduced. In conclusion, it effectively improves the efficiency and the effect of the payload dynamics parameters identification.