Many agents, one mission.
Real missions aren't flown by one machine. Work in this area studies coordinated autonomy across air, land, and sea: shared communication, distributed planning, and resilient behavior when links drop, agents fail, and the environment fights back.
Watch the explainer
The idea behind this lab in ninety seconds — then come back and drive it.
How the work happens.
The methods behind the research.
Teams, not soloists
Planning and control for fleets that divide tasks, cover area, and adapt as a group.
Coordinating over a lossy link
Behavior that degrades gracefully when bandwidth is scarce and the relay keeps changing, including who relays for whom.
No single point of failure
Decisions spread across the fleet, so the mission survives the loss of any one agent.
Air, land, and sea
Heterogeneous platforms sharing one picture of the mission and one communications fabric.
Open problems we're pursuing.
What's being pursued now.
Relay and hand-off
How a fleet keeps a connected mesh as agents move and links break, and proves it stayed connected.
Coverage under uncertainty
Coordinated search and survey when the map is wrong and the team is incomplete.
Resilient formations
Formations that reform around failure instead of falling apart.
The swarm, live on your GPU.
The actual WGSL quadrotor kernel from the swarm-bench, stepping a whole fleet in real time. The math is fixed-point, so the run is bit-exact and reproducible — the property a multi-agent mission needs before anyone can trust it.
Open the full view ↗The real fleet kernel on your GPU, not a canned animation: fixed-point and reproducible. An illustrative model of the multi-agent stack.
Work in this area.
Open positions, including the Physical AI Investigator Program, are listed on Careers. Research here can also spin out into a company.