ONBOARD SOFTWARE
Mission-specific autonomy that runs on the aircraft, not in a data centre.
What we build
Autonomy written for one problem at a time
Generic autonomy solves nobody's problem completely. We write mission-specific software that runs on the companion computer already on your aircraft — perception, decision-making, and data capture aimed at the job you actually need done. Built on open source foundations, delivered with source.
- ROS 1 & ROS 2
- PX4
- ArduPilot
- Micro XRCE-DDS
- Gazebo
- Docker
- C++
- Python
Mapping & Survey Noise mapping, terrain capture, and repeatable survey passes.
Survey work is a software problem more than a flying problem: the flight has to be repeatable, the data has to be georeferenced, and the second pass has to be comparable to the first. We build the onboard programs that make a mission repeatable and the output usable.
Noise mapping is a concrete example — flying a defined pattern while logging acoustic measurements against position, then producing a dataset that can be compared over time.
Programs we build
- Noise mapping and acoustic survey
- Repeatable automated flight patterns
- Georeferenced data capture
- Pass-to-pass change detection
- Onboard logging and export
Confined Space & Tunnel Scanning Flying and scanning where there is no GPS and no room for error.
Tunnels, shafts, and voids remove the two things most autonomy assumes: a GNSS fix and space to recover. We build onboard navigation and scanning programs that hold position from onboard sensing alone and keep the aircraft off the walls while it captures the geometry.
Programs we build
- Tunnel and shaft scanning
- GPS-denied navigation
- Onboard obstacle awareness in tight spaces
- LiDAR-based geometry capture
- Automated coverage patterns
Warehouse & Asset Inspection Indoor inventory, stock checks, and structural inspection routes.
Indoor inspection is the same problem as tunnel work with a commercial deadline attached. The aircraft has to navigate without GNSS, follow a route reliably enough to compare against last month, and hand back something a non-pilot can use.
Programs we build
- Warehouse inventory and stock checks
- Indoor route following without GNSS
- Structural and asset inspection
- Repeatable inspection passes
- Reporting output for non-pilot users
Perception & Detection Onboard recognition wired into what the aircraft does next.
Detection only matters if something happens because of it. We run perception on the companion computer and publish results into the autopilot over MAVLink, so a detection can hold the aircraft, re-route it, cue a gimbal, or escalate to the operator.
What we build
- Onboard inference on VOXL 2, Skynode, Jetson Orin
- Detections published over MAVLink
- Detection-triggered vehicle behaviour
- Model-agnostic pipeline
- Trained on your data
Something Else Entirely Most of this work starts as a problem nobody has productized.
The programs above all began as a customer describing a job their aircraft could not do yet. If your problem is not on this list, that is the normal starting point — not a reason it cannot be built.
How an engagement starts
- Feasibility assessment against your airframe
- Simulation in Gazebo before flight
- Surrogate vehicle testing
- End-to-end validation on your platform
- Source, configuration, and documentation delivered
These programs run on the compute and sensors covered under Sensor & Hardware Integration.
Have a problem your aircraft cannot solve yet?
Describe the mission and the airframe. We will tell you honestly whether it is a software problem, a hardware problem, or not yet possible.