A Distributed State Filtering Method Based on Square Root Volume Measurement Weighted Consistency

Abstract

The invention discloses a distributed state filtering method based on square root volume measurement weighted consistency, and relates to the problem of sensor network distributed state filtering in information fusion technology. The method first uses the square root volume rule to obtain the square root factor of each node's prediction information vector and information matrix of the target state; then calculates the corresponding local measurement information vector and measurement information matrix square root based on the local state prediction information and sensor node measurement information factor, and carry out weighted consistency iteration through the measurement information interaction between neighbor nodes; finally, update the state estimation of the target by weighting the state prediction information and consistency measurement information of each node. This method solves the problem of distributed nonlinear state filtering in sensor network well, improves the consistent convergence speed of each node's state estimation and the numerical stability of the method.

Description

technical field [0001] The invention relates to the problem of sensor network distributed state filtering in information fusion technology, and is applicable to various sensor network distributed target tracking systems. Background technique [0002] In recent years, sensor networks have been widely used in target tracking, environmental monitoring, wireless video networking and other fields. Compared with centralized state estimation technology, sensor networks have obvious advantages. The whole system has good scalability, small communication burden, and insensitivity to single node failure. Many advantages. [0003] Tracking the motion state of the target in the monitoring area is one of the basic tasks of the sensor network, and the research on the consistent state estimation method in the prior art is relatively extensive. For the linear Gaussian system, the estimation result close to the centralized method can be obtained by the average consistency method, and the non...

AI Technical Summary

Problems solved by technology

However, when the system is highly nonlinear, the method will be unstable and the estimation accuracy will be low

Compared with extended Kalman filtering, deterministic sampling methods such as insensitive filtering and volumetric filtering have better stability and higher estimation accuracy. However, due to the limited word length of the computer and the existence of truncation errors, it is difficult to guarantee the estimation error covariance time Symmetric positive definite, so that the square root operation of the error covariance matrix in the filtering process cannot be performed, causing the filter to fail

Especially for insensitive filters, the selection of scale parameters directly affects the final filtering performance. If the selection is improper, it is easy to cause filter divergence.

Although the particle filter can solve the problem of nonlinear state filtering very well, it has a large amount of calculation and relatively poor real-time performance, which makes it difficult to meet the requirements of engineering applications.

In addition, the existing consensus protocols often assume that all sensors in the network can observe the target, and the consistency rate factor between all neighboring nodes is also the same, resulting in slow convergence speed and relatively low accuracy of nodes in the entire network.

In practical applications, the number of sensing nodes in the network is often limited, and most of the nodes are communication nodes, which do not have the sensing function and are only responsible for forwarding the observation information of the sensing nodes.

In addition, the movement of the target and the observation of the sensor are often non-linear, and the computing power and energy of each node in the network are also limited. How to use limited resources to achieve distributed and effective estimation of the target state is a problem worthy of further study.

Images