Stability analysis of distributed Kalman filtering algorithm for stochastic regression model

TL;DR

This paper introduces a distributed Kalman filtering algorithm based on diffusion strategy for sensor networks, demonstrating its stability and cooperative estimation capabilities without a central fusion center.

Contribution

The paper presents a novel DKF algorithm that is robust, scalable, and stable under non-stationary conditions, enabling cooperative signal tracking in sensor networks.

Findings

The DKF algorithm is stable under non-independent, non-stationary conditions.

It enables cooperative estimation even when individual sensors cannot track the signal.

Simulation confirms the cooperative property of the proposed method.

Abstract

In this paper, a distributed Kalman filtering (DKF) algorithm is proposed based on a diffusion strategy, which is used to track an unknown signal process in sensor networks cooperatively. Unlike the centralized algorithms, no fusion center is need here, which implies that the DKF algorithm is more robust and scalable. Moreover, the stability of the DKF algorithm is established under non-independent and non-stationary signal conditions. The cooperative information condition used in the paper shows that even if any sensor cannot track the unknown signal individually, the DKF algorithm can be utilized to fulfill the estimation task in a cooperative way. Finally, we illustrate the cooperative property of the DKF algorithm by using a simulation example.