Feature identification for fuzzy logic based adaptive kalman filtering

Abstract

Standard approaches to fuzzy logic based adaptive Kalman filter use features based on adjustment of noise statistics according to performance of plant and sensor noise sources. Availability of this information is limited to specific domains. Here the Kalman gain computed by conventional Kalman filter is modified using online estimate of measurement residuals which is always available. Arguments are given for qualitative relationship between the residuals and Kalman gain tuning. This fuzzy logic based scheme is computationally simple and hence fast.