dronova.

Multi-agent systems

Autonomy for
drone teams.

Planning, mapping and control.
Tested in Python simulation and PX4.

Explore simulations Research
PX4 / SIHRecorded trajectory
PX4 prerequisite mission, oblique trajectory viewAll 452 recorded estimator samples, relative to the first position. Blue is the three-dimensional flight path; the faint line is its ground projection. Recorded duration 50.7 seconds. Axes are in metres.0246Altitude / mNorth / mEast / m5.83 m peak
01 / Prerequisite mission50.7 s / 452 samples
1,940Simulation episodesAcross four separate studies
4/4PX4 SIH missions passedIndependent simulated worlds
09USGS terrain sitesTrain / validation / test
07Survey plannersRecorded model-2.1 traces

Research areas

From control to coverage.

All five studies
USGS 3DEP / Normalized elevation

01 / Adaptive survey

Learning where to look.

Terrain-based mapping with policy selection, frozen baselines and held-out geography.

516 executions · 9 sites

02 / Control

Separation under constraints.

Projected control, actuator feasibility and independent swept-separation audits.

896 paired-comparison episodes

03 / Navigation

Planning around obstacles.

Three planners across five layouts. Waypoint completion, recovery and path-intersection analysis.

240 episodes · 1,440 agent missions

Original figures & experiment plots

Flight-stack integration

Native PX4.
Recorded end to end.

Takeoff. Waypoint. Hover. Return. Land.
Four accepted missions using the PX4 estimator and controller in software-in-the-loop.

TakeoffWaypointReturnLand
Mission results