Research theme
Learned communication
End-to-end learning of what to say, when, and to whom — under bandwidth, latency, and reliability constraints.
Communication in real multi-robot deployments is rarely free. We study how agents can learn compact, task-relevant message protocols that degrade gracefully when bandwidth drops or links are lost.
Recent papers
- Communication-constrained collective intelligence for indoor drone swarms ICRA 2026 · 2026
- Heterogeneous multi-robot reinforcement learning NeurIPS 2025 · 2025
People
- Amanda Prorok