Novel decoupling algorithm based on parallel voltage extreme learning machine (PV-ELM) for six-axis F/M sensors

Abstract

Abstract Accurate, time-effective calibration and decoupling procedures of multi-axis Force/Moment (F/M) sensors are critical to the implementation of these sensors. Recently, many decoupling methods have been proposed by researchers, but the inherent coupling relationship among components of multi-axis F/M sensors has not been taken into account. In this paper, we thus proposed a novel Sparse Voltage Maximum Inter-class Variance (SVMIV) algorithm, which took advantage of the inherent relationship nature. Furthermore, a novel nonlinear decoupling method based on the Parallel Voltage-Extreme Learning Machine (PV-ELM) was also presented. To demonstrate the utility of the proposed approach, extensive comparisons were made between the proposed PV-ELM decoupling method and several conventional decoupling approaches such as Least Square (LS), Support Vector Regression (SVR), Back Propagation Neural Network (BPNN), and Extreme Learning Machine (ELM). Results of real decoupling experiments demonstrated that the proposed the PV-ELM decoupling algorithm outperforms linear decoupling algorithms for six-axis F/M sensors. In addition, it was also proved that the PV-ELM decoupling method can decouple the outputs of six-axis F/M sensors with higher precision, faster speed, improved robustness, and faster convergence than the state-of-the-art nonlinear decoupling algorithms such as SVR, BP and ELM. Overall, this paper proposed a novel way to deal with the inherent coupling relationship of six-axis F/M sensors, and the experimental results demonstrated that the maximum I-type and II-type errors were below 0.356% and 0.270%, respectively, of full scale in all measured variables.