Solutions · On-board software
Pilot AI: AI-assisted autonomous pilot
Modular on-board software for autonomous navigation, decision-making and task execution by UAS and UGV platforms in dynamic environments. Developed following the European Space Agency’s ECSS standards for a more robust design and a resilient product.
Key features
Autonomy that runs on board
Real-time object detection
Retrained open-source deep-learning models, including thermal vision, running on the companion computer.
Adaptive path planning
Open-source libraries combined with our own implementations of state-of-the-art algorithms from recent research papers.
Autonomy modes
Fully autonomous operation, with teleoperation as a fallback.
Event-driven operations
Reaching a location or completing a navigation step can trigger actions: operating a payload, enabling an AI subsystem, and more.
Hardware-agnostic
Runs on MAVLink-based drones and rovers with PX4 or ArduPilot autopilots.
Observable and customisable
Algorithms are auditable and can be adapted to domain-specific needs.
Technical specifications
Built on a proven open-source stack
Open-source stack
MAVLink, PX4/ArduPilot and GNU/Linux.
Compute with AI
AI-capable companion SoCs for AI workloads.
Compute without heavy AI
Other SoCs for lighter workloads, or no companion computer at all.
Roadmap
In development
Navigation without GNSS or communications
Keeping navigation going when GNSS and communications are denied.
Alternative localisation
Localisation that does not rely on satellite navigation, supported by the ground station.
Dual use
Use cases
Civilian
Agricultural monitoring, infrastructure inspection, security surveillance and patrolling.
Defence
Autonomous operations, patrolling, surveillance, and moving to an objective to perform a task. More tasks can be developed on demand.
Tested with European agricultural SMEs
Prototype-validated and ready for external technical validation. Ask us for a technical deep-dive.