Networked Estimation using Sparsifying Basis Prediction

TL;DR

This paper introduces a networked state estimation framework that employs a sparsifying basis to reduce communication load, enabling accurate state recovery from limited transmitted information.

Contribution

It proposes a novel approach using a greedy algorithm to compute a sparsifying basis for efficient networked state estimation.

Findings

The framework effectively reduces data transmission requirements.

An upper bound for estimation error is established.

Numerical examples demonstrate the approach's viability.

Abstract

We present a framework for networked state estimation, where systems encode their (possibly high dimensional) state vectors using a mutually agreed basis between the system and the estimator (in a remote monitoring unit). The basis sparsifies the state vectors, i.e., it represents them using vectors with few non-zero components, and as a result, the systems might need to transmit only a fraction of the original information to be able to recover the non-zero components of the transformed state vector. Hence, the estimator can recover the state vector of the system from an under-determined linear set of equations. We use a greedy search algorithm to calculate the sparsifying basis. Then, we present an upper bound for the estimation error. Finally, we demonstrate the results on a numerical example.