Sensor Fusion Methods for Gaussian Mixture Models

TL;DR

This paper introduces decentralized consensus-based fusion methods for Gaussian mixture models in distributed sensor networks, enabling agents to collaboratively estimate target distributions without centralized coordination.

Contribution

It proposes novel fusion algorithms for Gaussian mixture models in distributed settings, handling both identical and differing prior estimates among agents.

Findings

Effective fusion of Gaussian mixtures achieved in distributed networks.

Algorithms ensure consistent posterior estimates across agents.

Applicable to scenarios with or without local observations.

Abstract

Consensus is a popular technique for distributed state estimation. This formulation allows networks of connected agents or sensors to exchange information about the distribution of a set of targets with their immediate neighbors without the need of a centralized node or layer. We present decentralized consensus-based fusion techniques for a system whose target prior estimates are a weighted mixture of Gaussian probability density functions (PDFs) for the following cases: 1) in which all agents have the same a priori Gaussian mixture estimate of the target, and 2) in which agents have different a priori Gaussian mixture estimates of the target. For the second case, we present a formulation that fuses each agent's a priori estimate without using local observations such that each agent's posterior estimate is the same across the network.