Volary works on autonomy that has to hold up outside the lab. The research picks a few hard problems and stays with them, and the platform turns that work into something reusable.
Deep focus on communication-limited, distributed, and degradation-prone problems. No trend-chasing.
Coordination that runs on-board across many agents, with no single point of control. This is the swarm-autonomy core, from formation and task allocation to decentralized decision making.
Most swarm work assumes clean links. Real deployments do not get that. Volary treats limited, lossy, and delayed communication as a first-class constraint in how agents plan and coordinate.
Autonomy that keeps working when a sensor drops, compute is tight, or conditions get messy. The question is not peak performance in simulation, but how the system fails and recovers in the field.
The same stack underpins every application track. Layered so a student can enter at any level, and a new payload or robot swaps in without rebuilding.
The physical platform: frame, motors, power.
IMU, cameras, depth, and the interfaces to the vehicle.
Attitude and rate control. PX4 as a reference implementation.
Visual-inertial odometry and localization for GPS-denied flight.
Turning raw sensor streams into obstacles, targets, and structure.
Building the spatial representation the planner reasons over.
Safe, dynamic trajectories. EGO-Planner as a reference implementation.
Getting from goal to behavior for a single vehicle.
Where Volary concentrates: many vehicles, limited communication, graceful degradation.
PX4 and EGO-Planner appear as reference implementations. Prior work, including FAST-Lab, is cited where it is used.
Individual results are fragile. We route them through a knowledge system so they survive past any one student or paper.
PhD work, experiments, and source material from partner programs and courses.
A structured store of concepts, methods, and engineering notes, kept separate from opinion.
A layered handbook, teaching material, and the stack that application tracks build on.
Autonomy claims mean little without repeatable flight. The platform track builds an indoor flight and validation area where the stack gets tested, application tracks get piloted, and students learn on real hardware.
The first application track is search-and-rescue. Agriculture and inspection follow as the platform and network grow.
Volary is open to research collaboration on swarm autonomy, communication-aware coordination, and field robustness, and to co-supervising or hosting students on the platform.