Partition-based Distributed Kalman Filter with plug and play features

TL;DR

This paper introduces a scalable, reconfigurable distributed Kalman filter for interconnected subsystems that accounts for dynamic coupling and uncertainty, with proven convergence and demonstrated effectiveness.

Contribution

A novel partition-based distributed Kalman filter that handles dynamic coupling, uncertainty, and allows plug-and-play reconfiguration with theoretical convergence guarantees.

Findings

Scalable online implementation with small matrix operations.

Effective handling of dynamic coupling and neighboring uncertainties.

Successful simulation validation of the proposed method.

Abstract

In this paper we propose a novel partition-based distributed state estimation scheme for non-overlapping subsystems based on Kalman filter. The estimation scheme is designed in order to account, in a rigorous fashion, for dynamic coupling terms between subsystems, and for the uncertainty related to the state estimates performed by the neighboring subsystems. The online implementation of the proposed estimation scheme is scalable, since it involves (i) small-scale matrix operations to be carried out by the estimator embedded in each subsystem and (ii) neighbor-to-neighbor transmission of a limited amount of data. We provide theoretical conditions ensuring the estimation convergence. Reconfigurability of the proposed estimation scheme is allowed in case of plug and play operations. Simulation tests are provided to illustrate the effectiveness of the proposed algorithm.