Fleet Coordination

Drones, legged robots, AMRs and vehicles — operating as one fleet

Efferax treats a site as a single system. Heterogeneous robots share a map, negotiate space, auction tasks, and hand work to each other — with every decision mirrored in one digital twin and every scenario reproducible in Unreal Engine 5.

FLEET COORDINATION • 6 UNITS • SHARED SEMANTIC MAP
Initializing fleet coordination viewport…
Agentic handoff: idle — fleet nominal
Zone: free
Live fleet state • p99 telemetry 14 ms
Waiting for fleet telemetry…
Coordination Principles

Six rules the fleet engine never breaks

One shared semantic world model

Every unit contributes to and consumes a single 3D semantic map: the drone fills in large-area geometry and thermal layers, ground robots refine it at close range. New obstacles seen by one robot are costmap updates for all.

Deadlock-free traffic management

Corridors and intersections are modelled as a time-expanded graph. Robots reserve space-time windows (VDA5050 zone sets) before entering; conflicts are resolved by priority, payload criticality and battery state — never by collision avoidance alone.

Market-based task allocation

Missions are decomposed into tasks and auctioned to the fleet. Bids account for distance, battery, payload fit and current commitments, so the right robot — not just the nearest — takes each job.

Air-to-ground handoffs

Aero-V surveys at altitude and flags anomalies; the agentic planner converts a flagged thermal ROI into a close-range inspection task for Quad-I or a contact task for Arm-M, with the drone dipping to confirm the hotspot during the scan.

Battery-aware scheduling

Charging, hot-swaps and drone pack changes are scheduled as first-class tasks so coverage never collapses. Degraded units gracefully hand missions to peers.

Safety is local, coordination is global

Fleet decisions are advisory; every robot still enforces its own 1 kHz Control Barrier Function envelope. Losing the fleet link degrades to safe autonomous behavior, never to a stop-the-world failure.

WORKED SCENARIO • THE ONE RUNNING IN THE VIEWER ABOVE

Air-to-ground thermal anomaly handoff

  1. 1Aero-V completes an overwatch lap and publishes a thermal layer with a 94.2°C ROI on Asset T-04.
  2. 2The agentic planner raises a HIGH-priority inspection task; the fleet auction assigns Quad-I (closest, 84% SOC, thermal payload).
  3. 3Quad-I diverts from patrol, reserves the corridor segment to T-04, and performs a 16-frame thermal burst + 2 mm 3D scan.
  4. 4Aero-V drops to 2.4 m to cross-validate from above; both scans are fused into the twin and a CRITICAL alert is published.
  5. 5Meanwhile AMR-01 and AMR-02 negotiate the shared zone via reservation — one holds, the other crosses — with zero human input.
  6. 6The operator sees the whole sequence replayable in the digital twin and the scenario is exported to Unreal for regression.
Fleet Digital Twin

What the fleet twin mirrors

Unit poses, velocities & headings
50–500 Hz per unit
Zone reservations & corridor graph state
event + 10 Hz
Task auction bids & assignments
event
Shared semantic map deltas
2–10 Hz
Anomaly events & handoffs
event
Battery, thermal & link health
1–5 Hz
Safety envelope violations (CBF)
event
Full mission replay
frame-accurate

Bring your own fleet mix

VDA5050 and Open-RMF adapters let Efferax coordinate third-party robots alongside ours.

Request a fleet simulation