Improved dynamic parameter identification method relying on proprioception for manipulators

• Published in: Nonlinear dynamics Vol. 105; no. 2; pp. 1373 - 1388
• Main Authors: Jia, Jidong, Zhang, Minglu, Li, Changle, Gao, Chunyan, Zang, Xizhe, Zhao, Jie
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.07.2021; Springer Nature B.V
• Subjects: Accuracy; Automotive Engineering; Classical Mechanics; Control; Dynamic models; Dynamical Systems; Engineering; Excitation; Identification methods; Maximum likelihood estimation; Mechanical Engineering; Model accuracy; Modelling; Multiple criterion; Nonlinear control; Optimization; Original Paper; Parameter estimation; Parameter identification; Proprioception; Robot arms; Robot control; Trajectory optimization; Vibration; Dynamics parameter identification; Excitation optimization; Dynamics; Flexible joint; Robotics; Model-based-control (MBC)
• DOI: 10.1007/s11071-021-06612-y
• ISSN: 0924-090X

Focusing on the complex nonlinear control problem of robots, this study concentrates on a dynamic modeling method to demonstrate the influence of excitation trajectory optimization and modeling accuracy of robots. The system has to be properly excited to improve the accuracy of the estimated parameters. An improved optimization method is proposed to deal with a series of influences caused by the under-constraint of a single criterion and over-constraint of multiple criteria. Proprioception perception is used for variable information to avoid the utilization of additional sensors, and parameter identification of the optimized excitation trajectory is performed based on the maximum likelihood estimation method. Simulation and experiment results show that the proposed optimization method can reasonably balance the disadvantages of single-criterion and multi-criterion optimization. A case study is also conducted to compare the proposed approach with other methods. The proposed approach can improve the accuracy of parameter estimation by at least 2.3% and reasonably reduce the drawbacks derived from single-criterion optimization. It can also effectively improve the anti-interference capability of identification parameters against noises, thus maintaining the accuracy of the estimated dynamic model.