GhostNav is a drop-in navigation module for existing drone airframes — replacing satellite dependency with resilient onboard positioning. The software stack is simulation-validated. The hardware module is in development.
Contested and congested electromagnetic environments are now standard operating conditions. Platforms that depend on GNSS lose precision navigation exactly when the mission needs it most.
Low-cost jammers deny GNSS signal across entire operational areas. Spoofing attacks feed false position data undetected — with no onboard warning.
IMU-only dead reckoning accumulates error fast. A platform can be tens of metres off target within sixty seconds of signal loss — and the error keeps growing.
Visual-inertial odometry reduces drift but still degrades over distance and in feature-poor terrain. Not enough for precision tasks in denied environments.
GhostNav integrates into existing drone platforms as a compact sensor and compute payload. No redesign. No new certification cycle. No vendor lock-in.
Camera and IMU data are synchronised and fused in real time — onboard, at flight speed. Continuously updated position estimate accurate through aggressive manoeuvres, vibration, and signal denial.
Before takeoff, the module builds a compact map of the mission corridor from satellite imagery. No live data connection required in the air. The map travels with the drone.
During flight, the onboard vision system matches what the camera sees against the pre-loaded map — anchoring position to real-world coordinates and preventing drift from accumulating over the mission.
NMEA GPS sentences over UART at 10Hz. Plug-compatible with PX4 and ArduPilot. The autopilot sees a GPS sensor. No firmware changes required.
Simulated 60-second GPS-denied flight leg over real satellite imagery. GhostNav sustains near-zero drift while IMU-only and VIO+IMU systems degrade past mission-usable accuracy.
The simulator runs the full GhostNav navigation stack against real satellite tile packs downloaded for your chosen location. Every result shown on this page was produced by the same pipeline.
Set launch point, target coordinates, drone speed, and GPS-denied zone. The simulator downloads real satellite tiles for that exact corridor — the same tiles the navigation stack will use in the air.
IMU-only, VIO+IMU, and GhostNav simulated simultaneously using real imagery and physics-accurate sensor models calibrated to the target hardware. No synthetic data.
Animated trajectory map shows divergence in real time. Error-over-time chart shows bounded versus unbounded drift. Full diagnostic artefacts saved per run.
The entire navigation stack runs on the module. No cloud dependency, no RF emissions, no ground station required. The flight controller receives standard GPS sentences over UART and never knows the difference.
Camera and IMU data synchronised and fused in real time — onboard, at flight speed. Accurate through aggressive manoeuvres, vibration, and signal denial.
Before takeoff, the module builds a compact map of the mission corridor from satellite imagery. No live connection required in the air. The map travels with the drone.
Onboard vision matches what the camera sees against the pre-loaded map — anchoring position to real-world coordinates and preventing drift from accumulating.
Simulation-validated against real satellite imagery across multiple geographies. The first hardware prototype integrates three proven components onto a single companion computer board, targeting the Holybro X500 v2 as the test platform.
Compute
8-core ARM + NPU. Real-time onboard processing at 1Hz for routes up to 100km. NPU headroom reserved for future algorithm upgrades without hardware change.
Vision
Global shutter, monochrome, VGA. Global shutter eliminates motion blur at drone velocities. The exact sensor the simulation was built around.
Inertial
Automotive-grade 6-DOF IMU, vibration-resistant. The noise profile behind every drift number shown in the results section.
Test Platform
PX4-native carbon fibre quadcopter. GPS active for calibration baseline. GPS disabled for navigation validation.
Complete
Navigation stack proven against real satellite imagery across multiple geographies. Bounded error demonstrated on structured terrain. Spoof rejection and graceful VIO fallback confirmed.
Complete
Full three-mode comparison running live against real tile packs. Route tile analysis confirms coverage, drift envelope handling, and self-correction across the full flight corridor.
In Progress
RK3588 + OV7251 + BMI088 bench integration. Camera-IMU calibration and latency profiling ahead of first flight.
Next
Calibration flights with GPS active to establish ground truth baseline. GPS-disabled validation to close the gap between simulation and production results.
Bolt the module to the flight controller stack using the included vibration-isolated mount.
Connect via UART to your autopilot. MAVLink telemetry is exposed automatically.
Run the guided calibration routine — under 10 minutes, no ground station required.
Model your mission profile in the Ghost Nav simulator, then deploy with validated confidence.
Run a live mission simulation — pick a location, set the GPS-denied zone, and compare GhostNav against IMU-only and VIO+IMU navigation in real time against real satellite imagery.
▶ Open the Simulator