Coordination & perception for heterogeneous autonomous fleets
One brain for every drone, robot & vehicle.
A modular, software-only coordination layer. Plug a module into any drone, ground robot, vehicle, camera, or sensor; Praemonitus fuses what they all perceive into one live, uncertainty-calibrated world model and coordinates them as a single fleet. Airframe-, model-, and form-factor-agnostic, built on ROS 2 and ArduPilot.
Built to make a mixed fleet operate as one
One module, any platform
A modular coordination layer that plugs into drones, ground robots, vehicles, fixed cameras, sensors, and comms devices alike. Built on ROS 2 and ArduPilot, each module adapts to what its platform can do — airframe-, model-, and form-factor-agnostic — so you add capability without re-architecting or locking into a single vendor.
One fused world model
Overlapping, noisy detections from every agent — air and ground, RGB and thermal, moving and fixed — merge into a single track list, each track carrying calibrated confidence. You see what is where, and exactly how sure the system is, across the entire fleet at once.
Coordinated, not just connected
Mission behavior-trees turn the shared world model into coordinated action: hand a track between a drone and a ground robot, re-acquire a target behind cover, task the right platform to the right job. Every agent works from one picture instead of its own stovepiped view.
Ask the live mission
A read-only AI copilot, powered by Claude, answers questions straight from the fused scene — “how many people have you found, and where?” It can propose a course of action; it cannot execute one. A human authorizes every action, and the boundary is enforced in code, not policy.
Resilient at the edge
Collaborative perception shares only high-value detections over a bandwidth-budgeted mesh and degrades gracefully when links drop. Run inference on local GPUs for low latency or burst to the cloud for heavier reasoning, behind a provider seam that falls back cleanly under a degraded link.
What Praemonitus is
Praemonitus is a modular, software-only coordination and perception layer that sits between an autonomous fleet and its operators. Plug a module into any platform — a drone, a ground robot, a vehicle, a fixed camera, a sensor, or a comms device — and it runs perception at the edge, then fuses those overlapping, noisy detections from every agent into a single, uncertainty-calibrated model of what is where. It is airframe-, model-, and form-factor-agnostic by design, building on open standards like ROS 2 for distributed communication and ArduPilot for flight control, so a heterogeneous fleet behaves as one coordinated system rather than a dozen stovepiped feeds.
Operators don’t just watch the world model — they act on it and can ask it. Mission behavior-trees coordinate and task agents across the fleet, while a read-only AI copilot, powered by Claude, answers questions grounded in the live fused scene and writes the after-action brief from the same audited mission log. The boundary is enforced in code, not policy: the copilot can propose a course of action, but it cannot execute one — a human authorizes every action. Inference runs on local GPUs for low latency or bursts to the cloud for heavier reasoning.
Designed for scalability and security, Praemonitus is a non-weapon, human-on-the-loop platform for civilian and dual-use missions — search-and-rescue, infrastructure inspection, security and perimeter awareness, environmental monitoring, and logistics. It turns complex multi-agent autonomy into something observable, auditable, and controllable in real time, across whatever mix of drones, robots, vehicles, and sensors the mission requires.
Frequently asked questions
What is Praemonitus?
Praemonitus is a modular, software-only coordination and perception layer for heterogeneous autonomous fleets. Plug a module into any drone, ground robot, vehicle, camera, or sensor and it fuses what they all perceive into one live, uncertainty-calibrated world model, then coordinates them as a single fleet. Airframe-, model-, and form-factor-agnostic, built on ROS 2 and ArduPilot.
How is it different from drone autonomy stacks like Skydio or Anduril?
Those stacks are airframe-coupled and built around one vendor's hardware. Praemonitus is the airframe-, model-, and form-factor-agnostic layer that sits across a mixed fleet — drones, ground robots, vehicles, cameras, sensors — and produces one coordinated world model with calibrated confidence. It is the coordination and fusion layer the hardware vendors structurally do not ship.
Can it coordinate a mix of drones, ground robots, and fixed sensors together?
Yes — that is the core design. Every platform runs the same perception seam and publishes into one shared world model over ROS 2, so a drone, a ground robot, and a fixed camera contribute to a single track list. Mission behavior-trees then task and hand off across them — for example, a drone spots a target and a ground robot is tasked to approach it — with a human authorizing every action.
Does the AI fly or control the drones?
No. The AI copilot is read-only and human-on-the-loop: it answers questions about the live fused scene and can propose actions, but a human authorizes every action. ArduPilot flies the aircraft, Nav2 drives the ground robots, and the operator decides.
What can I use Praemonitus for?
Civilian search-and-rescue (SAR), drone inspection of solar, wind and infrastructure assets, ISR and shared situational awareness, precision agriculture, and environmental monitoring.
Which platforms, robots, and sensors does it support?
It is airframe-, model-, and form-factor-agnostic. Drones (ArduPilot / MAVROS), ground robots (ROS 2 / Nav2 / MoveIt 2), vehicles, fixed cameras, sensors, and comms devices plug into the same coordination layer over ROS 2, with any mix of RGB and thermal cameras and perception models. New platforms plug in without re-architecting.
Stay in the loop
Join the early access list to follow progress, see live swarm demos, and hear when Praemonitus opens up to pilot partners.
Prefer email?