Drone LED Advertising Banners

We fly large indoor venues constantly. Whether it is a stadium halftime show or a convention center expo, the constraints are always the same: No GPS.

For years, the industry standard for indoor flight has been binary: either you invest $50,000 in an external motion capture system (Vicon/OptiTrack) for precision, or you rely on the pilot's skill to fly manually in "AltHold" mode. Optical Flow is an option, but over uniform stadium seating or low-light audience crowds, it is notoriously unreliable.

The "Blue Dot" Realization

During a recent deployment at a major stadium, I noticed something inconsistent. We were deep inside the concrete structure—a verified Faraday cage for GNSS signals. My flight controller showed 0 satellites, exactly as expected.

Yet, when I opened Google Maps on my phone to check a logistics pin, the "blue dot" was rock solid. It knew exactly where I was, down to the specific seating section. My phone had no GPS lock, but it had Network Fusion—a sophisticated algorithm blending Wi-Fi RTT (Round Trip Time), cellular triangulation, and IMU data.

The realization was immediate: The positioning data we need already exists. We just need to get it out of the Android OS and into the Flight Controller.

The Gap in Existing Tools

I assumed this was a solved problem. I loaded up Andruav, hoping to bridge the gap. The experience was disappointing; the connection stability wasn't there, and the UX felt experimental, not operational. I then turned to QGroundControl (QGC), the industry standard. I successfully connected it, but found a critical architectural limitation: QGC is designed to pass raw GPS observables to the drone. It explicitly ignores Android's "Fused" location provider because it lacks the satellite metadata QGC expects.

In short: The data was there, but no existing GCS would send it.

Engineering the Solution: MAVLink GPS

This led to the development of MAVLink GPS. We built a specialized bridge application that harvests high-precision Fused Location data from the Android framework and formats it into standard `GPS_INPUT` MAVLink messages.

By mimicking a U-Blox GPS module, we can trick the ArduPilot EKF3 into accepting this Wi-Fi/Cellular position as a valid navigation source. The result? Loiter mode indoors, without external beacons.

The Blimp Autopilot Synergy

This breakthrough is particularly critical for our internal Blimp Autopilot project. Unlike multirotors, blimps are large, high-drag airframes that drift significantly with even minor HVAC currents. Flying a blimp manually in a stadium requires constant, exhausting correction.

With MAVLink GPS, we can finally close the position loop indoors. The phone acts as the blimp's "absolute reference," allowing it to hold position over a crowd autonomously. This transforms the blimp from a high-workload manual platform into a set-and-forget aerial billboard.

We are releasing this tool to the community because we believe indoor flight shouldn't require a five-figure infrastructure budget. You can read the full technical breakdown in our Field Manual.