Trantor

Dynamic Task Assignment (DTA) is a foundational challenge in multi-robot systems: how do we efficiently allocate tasks to a decentralized swarm using only local information? In collaboration with Kleber Cabral and Arthur Freitas at the QUARRG Lab (Ingenuity Labs Research Institute), we developed a fully decentralized Task Assignment Mechanism (TAM) that mathematically guarantees 100% goal coverage.

The core of our approach allows heterogeneous robots to dynamically negotiate assignments in real-time, relying entirely on their localized sensing radius. By continuously sharing assignment states and known covered goals with neighboring units, the swarm collectively computes optimal assignments without a centralized controller, dynamically resolving conflicts as they approach their targets.

To validate the algorithm, we simulated complex real-world environments like dense warehouses, introducing arbitrary obstacles to block direct paths and limit line-of-sight communication. The results proved that regardless of the specific robot kinematics, underlying navigation policy, or obstacle density, our TAM guarantees complete goal coverage. This architecture provides a robust and highly scalable foundation for modern warehouse logistics, autonomous search and rescue, and industrial manufacturing swarms.