Abstract: Accurate dynamic models are crucial for the control and safety of collaborative robots in human–robot interaction tasks, and the core of this research is dynamic model parameter identification. Inaccurate modelling of friction terms in dynamic models can lead to poor robot control, stability, and motion accuracy. To address this issue, a modified Stribeck friction model is proposed, which adds acceleration and velocity terms to better fit the nonlinearities in the dynamic model. First, a robot dynamic model with joint friction terms is established using the Lagrangian method, and divided into linear and nonlinear components. Using a six-degree-of-freedom collaborative robot platform, third-order Fourier series signals are used as joint trajectory inputs. The collected data is processed and filtered to establish an experimental dataset for parameter identification. Finally, this dataset is used to identify the dynamic model parameters. The linear component is identified using the constrained iteratively reweighted least squares method, while the nonlinear component is identified using the density-based spatial clustering of applications with noise (DBSCAN) and the grey wolf optimization-cuckoo search hybrid optimization algorithm (GWO-CS). Experimental results show that the improved Stribeck friction model can accurately fit the nonlinear terms of joint friction. Compared with the traditional Coulomb–viscous friction model and the standard Stribeck model, the root mean square error (RMSE) of the joint torque is greatly reduced, effectively improving the parameter identification accuracy of the collaborative arm dynamics model.

Key words: collaborative robot, parameter identification, Stribeck friction, optimization algorithm