A Gaussian Particle Filter Approach for Sensors to Track Multiple Moving Targets

TL;DR

This paper introduces a Gaussian particle filter that combines Kalman and particle filtering techniques to estimate both the number and states of multiple moving targets from sensor data, even when the targets and associations are unknown.

Contribution

It proposes a novel Gaussian particle filter method that models each particle as a Gaussian distribution, improving target estimation in complex, uncertain scenarios.

Findings

Effective in estimating unknown number of targets

Handles non-Gaussian measurement models

Improves tracking accuracy in multi-target scenarios

Abstract

In a variety of problems, the number and state of multiple moving targets are unknown and are subject to be inferred from their measurements obtained by a sensor with limited sensing ability. This type of problems is raised in a variety of applications, including monitoring of endangered species, cleaning, and surveillance. Particle filters are widely used to estimate target state from its prior information and its measurements that recently become available, especially for the cases when the measurement model and the prior distribution of state of interest are non-Gaussian. However, the problem of estimating number of total targets and their state becomes intractable when the number of total targets and the measurement-target association are unknown. This paper presents a novel Gaussian particle filter technique that combines Kalman filter and particle filter for estimating the number and state of total targets based on the measurement obtained online. The estimation is represented by a set of weighted particles, different from classical particle filter, where each particle is a Gaussian distribution instead of a point mass.