New Algorithm Uses Hamilton's Rule for Robot Team Altruism
Inspired by ecology, a GNN policy optimizes robot team allocation for firefighting.
In a new paper on arXiv, researchers Riwa Karam, Ruoyu Lin, Brooks A. Butler, and Magnus Egerstedt tackle the challenge of heterogeneous multi-team robot collaboration. They propose a framework where robots are treated as transferable resources across teams, using Hamilton's rule—a principle from evolutionary ecology that governs altruistic behavior—to decide when a robot should help another team. The allocation problem is combinatorial and NP-hard, so the team developed a graph neural network (GNN) policy under a centralized training, decentralized execution (CTDE) paradigm. The GNN operates over a team interaction graph, predicting robot-level transfer decisions and reassignments based on the altruistic cost-benefit tradeoffs.
The approach was validated in a firefighting scenario, where multiple robot teams with different sensors, mobility, and firefighting capabilities must cooperate. Simulations and real experiments show the learned policy achieves near-optimal allocation performance while scaling gracefully to larger systems with more teams and robots. This work bridges ecology and multi-agent AI, offering a principled way to build collaborative robot swarms that balance selfish goals with system-wide efficiency. The results have implications for disaster response, warehouse logistics, and any domain requiring flexible, dynamic robot teaming.
- Framework treats robots as transferable resources and uses Hamilton's rule from ecology for altruistic decision-making.
- Graph neural network policy under CTDE solves NP-hard allocation problem with near-optimal performance.
- Validated in firefighting scenarios, demonstrating scalability to larger, heterogeneous robot teams.
Why It Matters
Enables efficient, scalable coordination of heterogeneous robot teams for complex real-world disasters.