Distributed Multi-Robot Active Gathering
Decentralized active perception and tracking in non-uniform environments.
Overview
This project studies distributed information gathering when targets are unevenly clustered. It reformulates regional search as a multi-armed bandit problem so that robots can balance exploration and exploitation without a central controller.
Method
The system combines distributed Bernoulli Thompson Sampling with PHD-based target estimation and Lloyd control. Robots share probabilistic information with teammates, while a goal-swapping strategy improves the assignment of search and tracking targets.
Publication
The work was published in Frontiers in Plant Science in 2025.