ADOS Platform
Software-defined drone intelligence
A full-stack open-source platform. Flight computer, onboard AI agent, ground control, vision engine, and cloud fleet management. One ecosystem. Any drone.
Architecture
Four layers. One platform.
Each layer works independently. Together, they make any drone autonomous.
ADOS Flight Computer
Onboard compute and real-time flight controller on a single board. AI inference, sensor fusion, motor control.
ADOS Drone Agent
Turns any Linux single-board computer into a drone brain. MAVLink proxy, video pipeline, cloud relay, suite runtime. 11 independent services.
ADOS Mission Control
Browser-based ground control station. Mission planning with 7 pattern generators, 38 configuration panels, gamepad flight control at 50Hz, 3D simulation.
DroneNet
Fleet telemetry and cloud management. MQTT real-time data, 4-tier drone integration model, fleet dashboard.
Communication
Two links. Full situational awareness.
Real-time drone command
- Telemetry protocol over serial between flight controller and compute module
- WebSocket relay to ground control station
- Gamepad and HOTAS input at 50Hz
- Sub-100ms latency in field mode
- Cloud fallback via MQTT at 200-500ms
Long-range HD video
- 5.8GHz broadcast protocol, not standard WiFi
- 50km+ range with off-the-shelf adapters
- 30-50ms glass-to-glass latency
- H.264 encoding on companion computer
- Ground station relays to any browser via WebRTC
AI Vision
See. Track. Act.
On-device AI vision with 31+ autonomous behaviors. Zero cloud dependency.
Real-time object detection on NPU. Person, vehicle, crop health, defects, thermal anomalies. Swappable models, no code changes.
Pixel-to-ground projection with Kalman-filtered tracking. Lock and follow targets at 30fps. Works in GPS-denied environments.
Follow-me, orbit, precision landing, obstacle avoidance, crop row following, structure inspection, quickshots, panorama, hyperlapse, mastershots.
Real-time image quality scoring. Coverage gap detection mid-flight. GSD calculation. Flag bad captures before you land.
Survey Pipeline
Capture. Process. Visualize.
End-to-end survey pipeline from aerial capture to 3D reconstruction. Your choice of processing tools.
On-device quality validation: blur, exposure, overlap scoring. EXIF and XMP geotagging compatible with commercial tools. RTK precision via NTRIP. Coverage tracking with gap detection.
Package datasets for WebODM, COLMAP, nerfstudio, or generic CSV. One click from the ground control station.
Plug into any tool. Open-source: WebODM, COLMAP, nerfstudio for gaussian splatting. Commercial: Pix4D, DJI Terra, Metashape. Not locked to any processor.
Coming soon: Studio tab in Mission Control. 100M+ point clouds with progressive LOD. 60fps gaussian splat rendering. Orthomosaics, elevation models, coverage analysis, PDF reports.
Not locked to any processing tool. Capture with ADOS. Process with anything.
Multi-Protocol
Speak every drone's language
One ground station controls ArduPilot, PX4, and Betaflight drones.
ArduPilot
Full Support83+ MAVLink decoders. 200+ parameters. 18 flight modes. 9 calibration types.
PX4
Full Support63+ parameter mappings. 18 modes. 3 PX4-specific configuration panels.
Betaflight
Full SupportFull MSP implementation. 34 decoders, 21 encoders. 8 Betaflight-specific panels.
Verticals
One drone. Six missions.
Software-defined verticals. Load a suite, transform the mission.
Sentry
Patrol and surveillance
Survey
Mapping and 3D reconstruction
Agriculture
Crop monitoring and precision spray
Cargo
Autonomous delivery
Inspection
Thermal and structural assessment
SAR
Search and rescue
Open Source
Built in the open
ADOS platform schematics, firmware, and software. All open-source. The community builds on our platform. Everyone wins.
Android didn't win mobile by building the best phone. It won by being the platform everyone built on. ADOS follows the same playbook: open-source creates the ecosystem, and the ecosystem creates the value.
Software Stack
- Flight Control FirmwareReal-time autopilot
- Onboard AI AgentPython, 11 services
- Ground Control StationTypeScript, 98K LOC
- Vision EngineNPU-accelerated computer vision
- HD Video Pipeline5.8GHz broadcast protocol
- Cloud Fleet PlatformMQTT and reactive database
Deployment
Deploy your way
Field Mode
Direct connection. MAVLink over serial or WiFi. Sub-100ms latency. No internet needed.
Cloud Mode
MQTT telemetry with reactive queries. Monitor from anywhere. 200-500ms latency.
Hybrid Mode
Both. Field control with cloud telemetry overlay. Best of both worlds.