01/ Problem
Problem
Coordinate multiple robots through cluttered environments with both centralized and decentralized search policies, and analyze where each wins.
← Back to projects
0022024
Multi-Robot MCTS
Centralized + decentralized planning across TurtleBot3 swarms.
- ROS 2 Foxy
- Gazebo 11
- PyTorch
- TurtleBot3
- Docker
02/ Approach
Approach
Monte Carlo Tree Search policy implementations in two flavors: a centralized planner with full observability, and a decentralized variant where each robot runs its own MCTS with shared goal posting. Distributed multi-agent DQN baseline for comparison.
03/ Hardware
Hardware
TurtleBot3 (simulated). Gazebo 11. ROS 2 Foxy. Containerized for reproducible runs.
04/ Result
Result
Documented behavior trade-offs across centralized vs decentralized regimes; baseline numbers and emergent-behavior analysis across adversarial agents.