Multi-sensor active fault-tolerant estimation method based on self-organizing Kalman filter

Abstract

The invention discloses a self-organizing Kalman filter-based multi-sensor active fault-tolerant estimation method. The method constructs a self-organizing Kalman filter structure, and realizes fault detection by using hard fault detection threshold, soft fault detection factor increase / decrease rate and change rate. And design compensation factors. At the same time, according to the accuracy of the sensing signals of each sensor, the information distribution coefficient is designed to realize the fusion optimal estimation and active fault-tolerant compensation among the sensors. The main steps include: building parallel Kalman filter subsystem and reference Kalman filter system; multi-sensor hard fault detection based on self-organizing Kalman filter; sensor soft fault detection based on soft fault factor increase and decrease and change rate, and realizing soft fault detection. Correction of faults; obtaining optimal estimate X(k). The self-organizing Kalman filter-based multi-sensor active fault-tolerant estimation method can effectively solve the redundant signal fault detection and active fault-tolerant problems of multi-sensor systems, and improve the fusion accuracy and fault-tolerant performance of multi-sensor systems.

Description

technical field [0001] The invention belongs to the field of fault-tolerant technical estimation of redundant sensor data in an active safety system, and relates to a self-organizing Kalman filter-based multi-sensor active fault-tolerant estimation method. Background technique [0002] With the rapid development of intelligent technology, the requirements for the perception ability and robust performance of multi-sensor systems are becoming more and more stringent. In order to ensure the reliability of the multi-sensor system, it is necessary to effectively analyze the fault-tolerant function of the system and the accuracy of the input signal, so that the accuracy and system reliability of the sensor acquisition signal can be effectively guaranteed after processing. [0003] Since the random signal has no established change law, its estimation cannot be completely accurate, and the so-called optimal estimation only refers to the optimum under a certain criterion. According ...

AI Technical Summary

Problems solved by technology

[0004] (1) The state dimension of the centralized Kalman filter is high, and the calculation burden is heavy. The "dimension disaster" brought by it makes the calculation burden increase sharply, which is not conducive to the real-time operation of the filter.

Dimensionality reduction filtering will lose filtering accuracy and even cause filtering divergence

[0005] (2) The fault tolerance performance of the centralized Kalman filter is not high, which is not conducive to fault detection

Faults in either sensor subsystem can pollute other states in the centralized filter, making the sensory information output by the combined sensor system unreliable

Images