Identification Method of Payload Dynamic Parameters of Industrial Robots Based on Adaptive Selection of the Optimal Experimental Data Samples

The precise quantification of payload dynamic parameters is of great significance for ensuring the movement control accuracy of the industrial robot. Thus, an identification method is proposed to quantify the payload dynamic parameters of industrial robots based on the adaptive selection of optimal experimental data samples. Firstly, the robot dynamic models in the loading and unloading conditions are established using the Newton-Euler recursive method, and the payload dynamic model is then derived by taking the difference between two models. Secondly, the influence of the payload on the joint torque of the robot is analyzed to determine the joint kinematic data required for payload dynamic parameters identification.The excitation trajectory of driving joint motion is then established based on the Fourier series, and an optimization objective function of the excitation trajectory is then formulated through the Hadamard inequality. Subsequently, the same excitation trajectory is used to drive the joint of industrial robot in the loading and unloading conditions, and joint angular displacement, angular velocity, angular acceleration, and joint torque are simultaneously collected. The collected data is subjected to low-pass filtering, and the filtered joint motion data is screened using the proposed optimal data sample selection strategy to obtain an optimal experimental data sample set. The selected data set is then input into the payload dynamic parameters identification model, and the weighted least squares method is employed to estimate the payload dynamic parameters. Finally, the proposed method is verified through experimental tests and comparisons with traditional identification methods. The results demonstrate that the proposed method maintains high identification accuracy across different load scenarios, and can reduce the amount of data samples by optimizing the sample selection, which provides a basis for enhancing the efficiency of online identification of load dynamic parameters.