PSO-GA-based excitation trajectory optimization for robot dynamics parameter identification

The grinding quality of industrial robots is limited by nonlinear dynamics, and conventional parameter acquisition methods often yield inaccurate models, affecting end-effector positioning. This paper proposes a modal-validation-based dynamic parameter identification strategy. Cross-validated finite element and experimental modal analyses determine frequency-domain constraints for excitation trajectory design, while the identification model remains a rigid-body model with joint friction. A PSO-GA algorithm optimizes the excitation trajectory to identify the minimum independent dynamic parameter set, including inertia tensor components, first moments of mass, link mass, and friction coefficients. The identified model is embedded into a controller. Trajectory tracking experiments show that compared with traditional parameters, the proposed method reduces RMSE by 73.2% and maximum error by 45.4%, demonstrating improved robot stability and accuracy.