NOTE-05 TECHNICAL NOTE
Inside the MAVLab Technical Paper
A full technical write-up of MAVLab v1.5.0 — architecture, methodology, and an honest evaluation.
MAVLab v1.5.0 is released and open-source. This note summarizes the technical paper that documents it.
📄 Read the technical paper on Zenodo — the DOI now points to the published version.
The Thesis
The barrier to learning drone systems is not conceptual difficulty — it is infrastructure. The paper argues for inverting the learning order: a protocol-faithful, zero-infrastructure simulator first, the professional ROS / Gazebo / SITL stack later.
Why It Matters
Lab Notes are Ascend Labs’ proof layer. A working prototype is evidence; a written-up, reproducible technical paper — with results we can defend — is stronger evidence. It shows we can build the thing and reason about it honestly.
What The Paper Covers
- Phone-first architecture — physics, autopilot, and a MAVLink server all run on one Android phone. No cloud, Docker, or bridge process. This is the design choice that makes MAVLab work in low-connectivity, low-hardware settings.
- Protocol-first methodology — we treated “will real QGroundControl accept MAVLab as a vehicle?” as the riskiest assumption and proved it before building depth. That sequencing is itself a contribution.
- The five surfaces — Cockpit, Controller, SIM (3D twin), Mission, and Ops — each mapped to a bootcamp learning step.
Current Evidence
The paper reports what is verified, and is explicit about what is not:
- Live QGroundControl acceptance passes on real devices, in both same-phone split-screen and desktop-over-Wi-Fi modes.
- The 3D digital-twin sustains ~60 FPS; 133 unit tests pass; the release APK is 43.9 MB.
- Released open-source (Apache 2.0), with a tagged v1.5.0 build.
What is not claimed yet: learning-outcome efficacy (the controlled classroom study is future work), and battery-per-hour / telemetry-latency figures (not yet measured).
What Comes Next
- Run the classroom study with the IEEE bootcamp cohort — pre/post concept assessment, and time-to-first-interaction versus a SITL baseline.
- Post the paper as a formal preprint (arXiv) and fold results back into MAVLab’s roadmap.
If you work in drone education, simulation, or field-readiness training, partner with Ascend Labs.
We report what is verified — and we are equally clear about what is still future work.
The riskiest assumption was never the physics. It was whether real ground-control software would treat a phone as a vehicle. So we proved that first.