Geometrically-informed sequential Monte Carlo method for dynamic swarm tracking
Tharani Rajapaksha1, Amirali Khodadadian Gostar1, Reza Hoseinnezhad1
1School of Engineering, RMIT University, Melbourne, Victoria, Australia.
ISA Transactions
|June 4, 2025
Summary
This study presents a new method for tracking autonomous agent swarms using their geometric properties. The approach enhances robustness against false alarms and accurately estimates swarm dynamics and formations, outperforming traditional filters.
Area of Science:
- Robotics and Autonomous Systems
- Control Theory
- Computer Vision
Background:
- Dynamic tracking of autonomous agent swarms is crucial for coordinated operations.
- Existing methods face challenges with sensor noise, false alarms, and dynamic formation changes.
- Treating swarms as single targets simplifies tracking but requires robust state estimation.
Purpose of the Study:
- To introduce a novel dynamic tracking approach for autonomous agent swarms.
- To leverage swarm geometric properties for enhanced tracking accuracy and robustness.
- To develop a method tolerant to high false alarm rates and dynamic formation changes.
Main Methods:
- Utilizing the sequential Monte Carlo method for state estimation.
- Characterizing swarm state by center location/velocity and geometric parameters.
- Formulating a novel likelihood function incorporating swarm geometry and false alarm tolerance.
Main Results:
- Accurate estimation of swarm movement, formation, and shape.
- Demonstrated robustness against high false alarms and missed detections.
- Outperformance of conventional particle filters in dynamic and challenging scenarios.
Conclusions:
- The proposed geometric-based tracking method offers superior performance for autonomous swarms.
- The approach effectively handles time-varying formations and sensor imperfections.
- This work advances swarm tracking capabilities in complex operational environments.
