Live swarm console • simulated data

Autonomy is the hard part. We fly it.

KESTREL is Ace Hacker's autonomous flight-intelligence engine. It perceives cluttered, GPS-denied environments onboard, plans collision-free paths in milliseconds, coordinates whole swarms with quantum optimization, and keeps a fleet safe and deconflicted in shared airspace.

Request a demo → Explore the engine

8 msperception to avoidance

256agents per controller

GPS−deniednavigation capable

kestrel://swarm/mesh-live IN FORMATION

swarm mesh • live

area coverage92.4%

Agents online

Mesh links

Avg battery

74%

Engine at a glance

Drones coordinated per controller

Swarm

0 ms

Perception-to-avoidance latency, onboard

Perceive

Faster multi-drone mission routing

Plan

Autonomous mission uptime

Operate

One engine, four instruments

From a single rotor to a thinking swarm.

KESTREL pairs each layer of autonomous flight with the right computation, deep learning to see and navigate a cluttered world onboard, quantum optimization to coordinate many drones at once, and hard real-time control that keeps every aircraft safe.

Perceive

Onboard perception

Detects obstacles, estimates depth, and localizes without GPS, all on the drone, in milliseconds.

detection • VIO / SLAM

Plan

Real-time planning

Plans smooth, collision-free trajectories through clutter and replans the instant the world changes.

trajectory opt • avoidance

Swarm

Swarm coordination

Allocates tasks and routes across hundreds of drones with quantum and hybrid solvers.

QAOA • task allocation

Operate

Fleet operations

Deconflicts airspace, manages energy and endurance, and runs missions end to end.

deconfliction • UTM

Perceive • Instrument 01

See the world, without asking a satellite.

Under a bridge, inside a warehouse, between buildings, GPS is gone and the drone is on its own. KESTREL perceives and localizes entirely onboard: it detects obstacles, estimates their distance, and knows where it is, from its own cameras and inertial sensors.

01
Visual-inertial odometryCamera and IMU are fused into a metric estimate of pose and velocity, so the drone tracks its own motion with no external reference.
02
SLAM with loop closureA live map is built as the drone flies; recognizing a place it has seen before corrects accumulated drift.
03
Detection and depthObstacles, people, and infrastructure are detected and their distance estimated, on the onboard accelerator, within milliseconds.

VIOSLAMmonocular depthedge inference

SLAM trajectory • estimated vs truthdrift 0.4%

flight path & loop closure

EstimatedGround truthLandmark

Detection confidence by classonboard

Depth • occupancy gridnear = brighter

Plan • Instrument 02

A safe path, recomputed every few milliseconds.

The world moves: a person steps out, a gust pushes a branch, another drone cuts across. KESTREL plans a smooth, dynamically-feasible trajectory through the obstacle field and replans continuously, so avoidance is not a reflex bolted on but the plan itself.

Trajectory through obstacle fieldreplanned live

planned path & alternatives

Chosen pathCandidatesObstacle

01
Kinodynamic planningTrajectories respect the drone's real dynamics, thrust, attitude, and velocity limits, so a plan is always flyable.
02
Continuous replanningThe trajectory is re-solved at high rate against the latest occupancy map, so a new obstacle is handled in milliseconds.
03
Safe fallbackIf no valid path exists, the drone slows and holds inside a guaranteed-safe envelope rather than gambling.

kinodynamicMPCoccupancy mapsafe set

Cost field • distance to obstaclespotential

Replanning latencyms per cycle

Swarm • Instrument 03

One controller, hundreds of drones.

Deciding which drone does which task, in what order, along which route, is combinatorial and explodes with fleet size. KESTREL casts swarm assignment and routing as optimization problems and solves them with quantum and hybrid methods, so a large fleet acts as one coordinated system.

01
Task allocationEvery task is matched to the best-placed drone under fuel, capability, and deadline constraints, a large assignment problem.
02
Multi-drone routingRoutes are optimized jointly for coverage and minimum total energy, cast as a QUBO and solved with QAOA.
03
Decentralized consensusDrones hold formation and share intent over a mesh, so the swarm stays coherent even if a link drops.

QAOAQUBOmesh consensuscoverage

Swarm area coverage92.4% covered

coverage & agent positions

UncoveredCoveredAgent

Task allocation • drones × tasksassignment

Routing solver convergencetotal energy

Formation consensus errorconverges to zero

Operate • Instrument 04

Many drones, one safe airspace.

A fleet is only as good as the operation around it. KESTREL deconflicts flight paths so no two aircraft claim the same volume at the same time, manages battery and endurance across the fleet, and runs missions from launch to recovery.

Airspace deconfliction0 conflicts

4D trajectories • separation enforced

Cleared routeResolved conflictCorridor

01
4D deconflictionEvery planned path is checked in space and time; conflicts are resolved by re-timing or re-routing before takeoff and in flight.
02
Energy-aware missionsBattery, wind, and payload are modelled so a drone is recalled or swapped before endurance becomes a risk.
03
UTM-readyKESTREL exposes standard interfaces for unmanned traffic management and integrates with existing airspace authorities.

4D deconflictionenergy modelUTM / remote ID

Fleet endurance • battery by dronestate of charge

Mission timelinephases by drone

Benchmarks • honest numbers

Where quantum earns its place, and where it does not.

We benchmark against strong classical baselines, tuned planners and heuristic schedulers. Learning wins on perception and per-drone planning; quantum optimization wins as swarm coordination grows combinatorial. KESTREL routes each job to the method that wins, and keeps the safety path classical.

Routing solve time vs swarm sizelower is better

Classical plannerKESTREL hybrid

Capability profilevs conventional autopilot

Architecture • sensor to swarm

Onboard where it must be, in the cloud where it helps.

Perception and planning run on each drone in hard real time; swarm optimization and fleet operations run on the ground or edge. Flight results and maps flow back so the whole system learns from every sortie.

Now flying with partners

Put a thinking swarm to work.

KESTREL is in early deployments in inspection, mapping, logistics, and security. If you fly drones at scale, or want to, we will run the engine on your platform and your mission.

Request a demo → Read the technical brief

Runs on your airframe • edge or cloud • your flight data stays yours