Autonomous Tracking of Intermittent RF Source Using a UAV Swarm

TL;DR

This paper presents an autonomous UAV swarm system for localizing and tracking intermittent RF sources, comparing different algorithms for estimation and path planning to optimize accuracy and computational efficiency.

Contribution

It introduces a combined approach of recursive Bayesian estimation and steepest descent path planning for UAV-based RF source tracking, demonstrating superior performance over other methods.

Findings

Steepest descent path planning outperforms bio-inspired heuristic by an order of magnitude.

Recursive Bayesian estimator slightly outperforms detection-based EKF.

The system achieves accurate localization with efficient computation.

Abstract

Localization of a radio frequency (RF) transmitter with intermittent transmissions is considered via a group of unmanned aerial vehicles (UAVs) equipped with omnidirectional received signal strength (RSS) sensors. This group embarks on an autonomous patrol to localize and track the target with a specified accuracy, as quickly as possible. The challenge can be decomposed into two stages: 1) estimation of the target position given previous measurements (localization), and 2) planning the future trajectory of the tracking UAVs to get lower expected localization error given current estimation (path planning). For each stage we compare two algorithms in terms of performance and computational load. For the localization stage, we compare a detection based extended Kalman filter (EKF) and a recursive Bayesian estimator. For the path planning stage, we compare steepest descent posterior Cramer-Rao lower bound (CRLB) path planning and a bio-inspired heuristic path planning. Our results show that the steepest descent path planning outperforms the bio-inspired path planning by an order of magnitude, and recursive Bayesian estimator narrowly outperforms detection based EKF.