Information-Weighted Consensus Filter with Partial Information Exchange

TL;DR

This paper introduces an information-weighted consensus filter that reduces communication bandwidth in sensor networks while ensuring convergence to the centralized Kalman filter, validated through theoretical proofs and simulations.

Contribution

It proposes a novel partial information exchange method in consensus filtering that maintains convergence to the centralized Kalman filter while reducing bandwidth.

Findings

The algorithm guarantees convergence to the centralized Kalman filter.

Bandwidth is effectively reduced through partial information exchange.

Theoretical stability and convergence are mathematically proven.

Abstract

In this paper, the information-weighted consensus filter (ICF) with partial information exchange is proposed to reduce the bandwidth of the signals transmitted between the sensor nodes and guarantee its convergence to the centralized Kalman filter (CKF). In the proposed algorithm, a part of information chosen with the entry selection matrix is transmitted to the sensor nodes in the neighborhood at each consensus step, and consensus averaging is conducted at each sensor node with the partial and the local information. This ensures that the proposed distributed estimation algorithm converges to the centralized algorithm, while allowing the proposed algorithm to achieve bandwidth reduction of the signals transmitted between the sensors. With the proposed algorithm, the stability of the estimation error dynamics is proven and the convergence to the centralized algorithm is mathematically shown using the property of the average consensus. Simulations are conducted to validate the proposed ICF with partial information exchange and the related theoretical findings.