Predicting MEMS gyroscope's random drifts using LSSVM optimized by modified PSO

Abstract

A predictive modeling method for random drift of MEMS gyroscope is proposed, which is based on the least squares support vector machine optimized by modified particle swarm algorithm. We built a forecasting model of MEMS gyroscope's random drifts with least squares support vector machine, then used the modified particle swarm algorithm to optimize the model. The optimized LSSVM model was then adopted for prediction of MEMS gyroscope's random drifts. The modeling method solves SVM's disadvantage of slow training speed and requesting more resources. In addition, the modified PSO is more suitable for selecting global or local searching capability. The experimental result demonstrates the modeling method can effectively predict MEMS gyroscope's random drifts, and is more appropriate than LSSVM optimized by PSO.