Selected research · 2019—2022
Research is a coordination problem.
Four investigations into the structures that let intelligent systems exchange information, reason about one another, and operate safely at scale.
Selected case studies
Emergent discrete message communication
A compact, interpretable vocabulary learned by agents through broadcast, listening, and self-attention.
Deep implicit coordination graphs
A learned dynamic graph discovers which agent interactions matter before graph reasoning turns structure into action.
Urban air mobility at system scale
Simulation-driven analysis reveals which fleet, infrastructure, and traffic policies control capacity and throughput.
Collision avoidance in dense airspace
A deep reinforcement learning correction layer helps traditional avoidance logic remain safe and efficient as traffic density rises.
Earlier experiments span distributed training, driver uncertainty, semantic segmentation, and autonomous highway behavior.
Visit the original archive ↗