Primal–dual interior point strategy based parameter identification method for serial industrial robots

To accurately identify robot dynamics parameters and improve control performance, this paper proposes a full-parameter identification method based on primal–dual interior-point optimization (FPI-PDIP). Firstly, linearization of the robot dynamics equation combined with QR decomposition is adopted to extract the minimal set of inertial parameters (MSIP). Physical feasibility constraints are constructed according to the intrinsic physical characteristics of inertial parameters, transforming the identification task into a constrained optimization problem solved via the primal–dual interior-point method to acquire physically feasible full dynamics parameters. Optimized Fourier-series excitation trajectories and second-order Butterworth filtering are adopted to guarantee high-quality measurement data. The performance of FPI-PDIP is verified through simulations on a 7-Degree-of-Freedom (7-DOF) manipulator and physical experiments on a SCARA robot. Compared with the conventional MSIP scheme, the mean root-mean-square-error (Mean RMSE) of joint torque estimation is reduced by 46.09% across four joints of the 7-DOF robot in simulation. For the two joints of the SCARA robot, the joint-averaged Mean RMSE decreases by 53.59% on identification trajectories and 40.85% on unseen validation trajectories. The results verify the high accuracy and favorable generalization performance of the proposed FPI-PDIP method for robot dynamics feedforward control.