/technology / autonomy-stack
How vision becomes an autonomous mission.
The credibility centrepiece — the real perception, control, and mission pipeline.
Perception
Visual-inertial odometry.
A camera-and-IMU pipeline fuses tracked visual features with inertial measurements into a continuous pose estimate — the foundation everything above it depends on.
Perception payload
Control
PX4 integration.
The estimate feeds PX4 through a defined integration path, so autonomy rides on a mature, well-understood flight stack rather than a bespoke controller.
Services
ROS 2 services & health monitoring.
Mission and safety services run on ROS 2. Health monitoring watches the system continuously and drives the degraded-mode logic below.
Safety
Degraded modes & mission control.
When a subsystem degrades, the drone is designed to hold, return, or land rather than push on. Mission control sequences the flight and enforces these boundaries.
Validation
Offline operation, logs, and sim-first workflow.
Missions run offline by design. Every flight produces logs that are reviewed against expected behaviour.
Validation is sim-first: mission logic is exercised in simulation before it flies. Screenshots from the real stack are published here as they become presentable.
See it fly.
Claims here are scoped to what has actually flown.