A Fault-tolerant Data Fusion Method Based on Decentralized Kalman Filter for Redundant Sensor Configurations

Abstract

In this paper, we present a decentralized data fusion architecture to achieve Fault Diagnosis (FD) and Fault Tolerant Control (FTC) of sensors, which is effective under circumstances as control systems with redundant sensors, such as large unmanned helicopters. In the proposed fusion architecture, data from each sensor is firstly performed a novel FD based on Wavelet Transform (WT) and multi-sensor comprehensive analysis. Then it will be integrated if no fault detected. Structure of the fusion method is a two-layer decentralized filter: primary data are processed by decentralized Kalman Filter to attain more accurate results and descriptions of precisions respectively in local fusion nodes; the central fusion node integrates data from local fusion nodes based on a fast Covariance Intersection (CI) algorithm to obtain an optimal estimation of the measured state. The data fusion architecture is validated by actual flight data and fault injection from a large-scale unmanned helicopter. Results of the experiment indicate that the proposed fusion architecture gives remarkable enhancement to fault tolerance and precision.