Flight · SAPIENCE · July 2026

Flying with visual-inertial odometry

An experimental research flight from SAPIENCE development in July 2026: our first autonomous flight with visual-inertial odometry as the estimator's only source of odometry.

Organization: University of Alabama in Huntsville
My contribution: Robotics software and flight operations, from the ground station

Recorded EKF2 estimate and GNSS position traces, plotted at the same scale. — Experimental research flight, July 2026. Solid blue: EKF2 estimate from VIO alone. Dashed gray: GPS, logged as ground truth, not fused.

The aircraft and flight

Our team at UAH built the aircraft from off-the-shelf and printed parts. It ran PX4 on a Pixhawk 6X, with a Jetson Orin Nano onboard. I worked on robotics software and took part in flight operations from the ground station, deploying and monitoring the stack.

This was an experimental research flight, part of normal development, and our first autonomous flight with VIO as the only source of odometry for PX4's estimator. GPS was logged as ground truth alongside the EKF2 estimate and wasn't fused, so both traces are available to look at.

Reading the traces

Both traces use the same scale. The long straight segment near the end is one jump in the estimate between two log samples. This is one flight log, so it isn’t a general measure of navigation accuracy.

The positions are translated and rotated, so the file shows the track’s shape only. The map in the interactive version is an illustration.

Replay the flight Download flight data (JSON)

This flight was part of UAH’s work on SAPIENCE. For more on the project itself, see sapienceproject.com.