What is TensorFleet?
Developers building autonomous systems often get a single robot working, then stall when it’s time to scale. Multi-robot coordination, health monitoring, over-the-air updates, and safety fallbacks multiply the surface area fast.
TensorFleet is an SDK that helps you go from one drone/robot to production fleets with the same codebase — providing fleet APIs, mission graphs, and ROS/PX4 adapters so you can ship faster with confidence.
Why TensorFleet
- Ship features, not glue code — skip writing bespoke coordination services and control planes.
- Scale with confidence — the same abstractions work for one, ten, or hundreds of robots.
- Interop with what you use today — ROS 2, PX4, and your existing autonomy stack.
Core Concepts
1) Fleet APIs
A typed API surface to register vehicles, assign missions, stream telemetry, and manage updates. Build dashboards, ops tools, or CI hooks without re-inventing the backend.
2) Mission Graphs
Declarative DAGs for multi-step operations (e.g., takeoff → waypoint sweep → capture → return → upload). Reuse graph templates across sites and vehicles; inject parameters (geofences, waypoints, payload settings) per mission.
3) Adapters (ROS / PX4)
Drop-in adapters to speak your robot’s language. Use existing ROS 2 nodes for perception/planning, or PX4 for flight control, while TensorFleet handles orchestration and fleet state.
How It Fits Your Stack
- Your Autonomy (planning/perception/local control)
- TensorFleet Runtime (mission graphs, fleet control, health, updates)
- Adapters (ROS 2 / PX4)
- Your UI & Ops (operations console, alerts, analytics)
Keep your autonomy where it is. Add TensorFleet around it to handle fleet-grade orchestration.
Quick Start (Conceptual)
- Register vehicles — add robots to the fleet registry with metadata (type, capabilities, firmware).
- Define a mission graph — compose steps like
Arm → Takeoff → SurveyArea → Capture → Land. - Dispatch — assign missions with parameters (AOI polygon, altitude, speed).
- Monitor — stream telemetry/logs; auto-pause/resume; receive anomaly alerts.
- Review & iterate — store artifacts (images, logs), update templates, roll out OTA updates gradually.
Example Mission Graph (Pseudo)
Use the same graph across sites; swap parameters (polygon, altitude, payload mode) per run.
Common Use Cases
- Industrial Inspection — autonomous surveys, defect imagery, automatic uploads
- Logistics & Inventory — yard/warehouse counts, dock flow analysis
- Public Safety & Utilities — corridor inspection, rapid situational awareness
- Research & Pilots — go from single-demo to multi-site trials with one SDK
What You Don’t Have to Build
| Problem | TensorFleet Handles |
|---|---|
| Mission orchestration | Mission graphs, retries, branches, cancel/abort |
| Fleet registry & auth | Vehicle identities, roles, tokens |
| Telemetry pipelines | Streams, storage hooks, downsampling |
| OTA & configuration | Versioning, staged rollouts, rollback |
| Health & safety | Heartbeats, geofences, failsafes |
FAQ
Do I have to replace my autonomy stack?
No. TensorFleet wraps your stack with fleet-level orchestration and integrates via ROS/PX4 adapters.
Does this work for ground robots or boats?
Yes. The abstractions are vehicle-agnostic (air/ground/sea).
Is there a dashboard?
You can build one with the Fleet APIs, or start from reference UIs.
Get Early Access
We’re onboarding early teams who are scaling from single robot to multi-site fleets. If you’re ready to move beyond demos and pilots, join the waitlist and we’ll help you stand up your first fleet in days — not quarters.