Research Status and Trend of Decoupling Algorithms for Multi-dimensional Force Sensors

Abstract

Driven by the strategy of independent development of high-end equipment, multi-dimensional force sensors, as the core perception units in scenarios such as precise assembly of robots and in-orbit docking of spacecraft, have their measurement accuracy severely restricted by the structural coupling of elastic bodies and the coupling of manufacturing process errors. The comprehensive error of undecoupled six-dimensional sensors is relatively large. Decoupling algorithms have undergone three generations of development. Early linear methods were computationally efficient but unable to handle nonlinear coupling. In the intelligent algorithm stage, neural networks, support vector machines, etc. are introduced to improve the accuracy; The new paradigm of dynamic robustness enhances industrial adaptability. At present, there are still challenges such as real-time performance, calibration cost and consistency in mass production, and robustness against environmental interference. This paper analyzes the failure mechanism of traditional linear models, compares the advantages of intelligent methods, elaborates on innovative anti-interference technologies, constructs a multi-dimensional evaluation system, and looks forward to the integrated development directions such as deep learning for lightweighting, self-decoupling structures of piezoelectric materials, and adaptive calibration of digital twins, providing theoretical support for applications in extreme working conditions and promoting technological progress in high-end equipment.