Robot Dynamics Modeling with a Novel Friction Model and Extracted Feasible Parameters Using Constrained Differential Evolution

Focusing on the problem of extracting a set of feasible parameters to characterize the standard Newton-Euler (SN-E) dynamics model of robots, as an alternative to the linear matrix inequalities framework, a composite multiobjective differential evolution (MODE) algorithm based on the constraints of the linear combination vector inferred from the dynamic variables and the physical characteristics of the rigid links is proposed to recover feasible parameters from the estimated set of basic inertial parameters. In order to solve the challenge of difficult population evolution caused by complex feasible regions, the trial vector generation strategy for convergence is implemented through the elaborately combination of feasibility rule and ε constrained method. In addition, the peak deviation of prediction of the classical viscous-coulomb (Vis-Cou) friction model during the robot joint inversion has thus far remained an open question, which is addressed by a specially developed improved model based on the viscous-arctan function. The proposed solutions have been validated with a set of experiments on a six degree-of-freedom (DOF) robot. The experiment results show that the established SN-E model tracks the actual torque effectively, and compared with the classical friction model, the optimized model reduces the deviation of the predicted torque by 22.1