A Machine Learning-Based Compensation Method for Inductosyn Angle Measurement Errors
Xinyi Wang, Youtao Shen, Shenzhi Wang, Changbo Ma, Guangcheng Ma, Hongwei Xia · IFAC-PapersOnLine · 2025
To address the challenge that traditional inductosyn angle measurement error compensation methods struggle to identify and correct higher-order unmodeled components, this paper proposes a machine learning-based compensation approach. First, based on the operational principles of the inductosyn, the sources of error are analyzed and mathematically modeled. Then, a support vector regression (SVR)-based algorithm is introduced to identify and compensate for error components. Finally, experimental validation is conducted using a physical system. The results demonstrate that for an inductosyn with an initial mean squared angle measurement error of 1.37 degrees, the conventional Ordinary Least Square method reduces the error to only 0.0271 degrees, whereas SVR achieves a compensation accuracy of 0.0005 degrees with minimal impact on real-time performance. This represents an 82% improvement over traditional compensation methods, highlighting the proposed approach’s feasibility and superiority.