ArduPlane mission playback
A recorded SITL mission and telemetry can be inspected in the research viewer.
RECORDED SIMULATION This is not a physical flight. PROJECT 02 / AEROSPACE & AI RESEARCH LOG · 2026
AN EXPERIMENT IN ACCESSIBLE INTELLIGENCE
Make flight
intelligence
accessible.
Exploring how reused smartphone computing can help bring perception and AI capabilities to more affordable intelligent aircraft.
VISION / PERCEPTION
MOBILE COMPUTE
CONCEPTUAL AIRFRAME VISUALIZATION ≠ FLIGHT TEST
001 / SYSTEM STUDY
N 35.0° / CONCEPT
01 / WHY IT MATTERS
“The intelligence inside an aircraft shouldn't be determined only by what someone can afford.”
02 / ENGINEERING APPROACHARCHITECTURE / PROPOSED
A second life
for everyday compute.
Smartphones already contain processors, cameras, and sensing hardware. This project explores how that computing capability could be repurposed alongside dedicated flight-control hardware — rather than replacing critical real-time control with an experimental device.
A / 01
Reused mobile compute
Explore existing phone processors, cameras, and sensors for high-level perception tasks.
DATA / COMMANDS↔
B / 02
Dedicated flight control
Keep safety-critical stabilization on suitable flight-control hardware, with defined interfaces.
CONTROL OUTPUT→
C / 03
Fixed-wing platform
Evaluate the integration through controlled tests before claiming practical autonomous flight.
ⓘ
Design intent, not a completed integration claim. The diagram explains the proposed research direction. The recorded tests below are separate evidence and do not establish onboard phone-to-flight-controller integration.
03 / RECORDED TESTSINTERACTIVE RESEARCH VIEWER
Real test records.
Inspectable, not embellished.
Explore a recorded ArduPlane software-in-the-loop mission, the original 72-case environment campaign, and an independent Android camera-and-IMU capture test. This viewer is read-only and does not connect to or command an aircraft.
04 / WHAT THE EVIDENCE SHOWSVALIDATION STATUS
Separate what is tested
from what is still a goal.
This project combines several kinds of work. The distinctions matter: a simulation, a device-level experiment, and an integrated aircraft test are not interchangeable.
72-case test matrix
The recorded campaign contains 29 normal passes, 35 normal failures, and 8 guard-pass-only cases.
RECORDED SOFTWARE TESTS These results describe the test campaign, not autonomous flight performance.Android camera & IMU capture
An independent Android device test explored camera and motion-sensor capture.
DEVICE-LEVEL TEST It was not integrated with the flight controller.Phone-assisted aircraft intelligence
The wider architecture remains an exploration: use accessible mobile compute for higher-level perception while retaining suitable dedicated hardware for safety-critical control.
NOT YET VALIDATED AS AN INTEGRATED SYSTEM Future work must verify interfaces, latency, reliability, and safe operating limits.05 / WHAT COMES NEXTRESEARCH DIRECTION