Smartphone computing
Independent Android camera and IMU capture observed on a physical phone.
HARDWARE CAPTURE OBSERVEDENGINEERING STUDY / 01
Exploring how a smartphone can support a fixed-wing flight controller with affordable sensing and computation.
Watch the recorded flight ↗Simulation evidence.
Independent phone capture.
Future onboard integration.
03 / RECORDED SIMULATION
Follow a complete recorded simulation mission from takeoff through return. The path and aircraft position come from ArduPlane telemetry; the aircraft shape is illustrative.
The dashed path shows all recorded valid positions; the gold trail shows progress. The aircraft silhouette is conceptual. Position is a local projection, not a navigation map. Gaps represent unavailable data.
01 / THE TEST CAMPAIGN
Gazebo, ArduPlane and visual processing were exercised across five environmental factors. This is the original October 20 campaign, including its failures.
64 normal combinations vary image noise, blur, visual-inertial IMU noise, wind and CPU inference load across two seeds. Eight additional cases evaluate fault protection. Select a case to inspect its recorded outcome and conditions.
FAILURE ANALYSIS
The original review classified the 35 normal failures as visual quality (10), wall-clock real-time performance (6), external-source duration (4), and anchor geometry before authority (15). The preserved campaign is not an all-pass result. Candidate repairs and later development reruns have their own scope and do not rewrite it.
The original case outcomes are preserved in the campaign data. This viewer inspects those recorded results; it does not rerun the simulator.
Recorded categories for the 35 normal failures in the original campaign. Guard-pass cases are not counted as normal mission successes.
02 / ANDROID PROTOTYPE
A separate Android application was installed and operated on a physical phone. This was an independent sensor capture, not a connected aircraft flight.
DEVICE CAPTURE / OCT 2026
A separate Android test recorded 3,307 camera frames with zero reported dropped images or records. The accelerometer and gyroscope were each observed at approximately 96.9 Hz.
The phone AI and native runtime were not verified in this test. The capture therefore demonstrates sensor collection only, not onboard AI processing or flight-controller integration.
Independent camera and IMU capture only · No physical flight-controller closed loop
04 / SYSTEM CONCEPT
The architecture separates evidence already gathered from the integration and autonomy work still ahead.
Independent Android camera and IMU capture observed on a physical phone.
HARDWARE CAPTURE OBSERVEDArduPlane telemetry and modes explored in software-in-the-loop simulation.
SIMULATION OBSERVEDCamera inputs and AI-assisted perception remain part of the proposed onboard system.
INTEGRATION IN PROGRESSNavigation assistance, tracking, obstacle avoidance and return require further integration and validation.
FUTURE CAPABILITY05 / EVIDENCE & LIMITS
This presentation uses a recorded SITL telemetry stream and the original project’s replay position utilities. It does not connect to an aircraft or send commands.
Phone camera and sensor capture was tested independently. A complete phone-to-flight-controller system and autonomous real-world flight have not been demonstrated by the evidence in this demo.
Evidence summary · recorded SITL and independent Android capture