Xwing CTO Talks Software Stack for Autonomous Cargo Aircraft

A Q&A with the CTO of Xwing, a California-based aviation startup looking to develop the next generation of autonomous air cargo aircraft.

Xwing’s software stack and mission control center completed what the company claims is the world’s first fully autonomous gate-to-gate flight of a commercial cargo aircraft, a Cessna Grand Caravan 208B, during a February 2021 flight remotely controlled from their mission control center in Concord, California.

In February, Xwing completed an autonomous air cargo aircraft takeoff, navigation and landing in San Francisco, remotely piloting the aircraft from its mission control center in Concord, California.Xwing

After spending four years in stealth, Xwing emerged as a next-generation aerospace technology supplier in May 2020, with executive leadership consisting of former Terrafugia co-founder Anna Dietrich and Lockheed Martin veteran Ed Lim among others to make pilotless flight of small passenger aircraft a reality.

"Over the past year, our team has made significant advancements in extending and refining our AutoFlight system to seamlessly integrate ground taxiing, take-offs, landings, and flight operations, all supervised from our mission control center via redundant data links," Marc Piette, CEO and founder of Xwing said in an April 15 press release. "Additionally, our piloted commercial cargo operations have been delivering critical supplies including COVID-19 vaccines, to remote communities since December 2020.”

Instead of building autonomous airplanes, the San Francisco-based aviation startup wants to use its agnostic software stack to transform manned and unmanned aircraft flight hardware into systems capable of navigating from takeoff to landing. Their ambitious goal has proven viable enough to draw investment from Thales and several venture capital firms including Array VC, ENIAC Ventures, and R7 Partners among others to the tune of $14 million across its first two funding rounds.

Since launching operations in May, Xwing has completed non-commercial flight demonstrations of the first fully autonomous air cargo aircraft. Their most recent hire will bring Jesse Kallman, a former Airbus executive, to their leadership team. Keith Allen, the former Chief Engineer for flight controls on the AACUS (Autonomous Aerial Cargo Utility System) UH-1 "Huey" helicopter program and Aurora's 11,000-pound Orion UAV has been hired to lead technical engineering development.

Below, check out our question and answer session with Maxime Gariel, CTO of Xwing and former Collins Aerospace flight control systems engineer about the February test flight and what’s coming up next for Xwing.

Maxime Gariel is the CTO of Xwing.


Avionics: How did Xwing modify the Cessna Grand Caravan in a way that enabled the company to remotely pilot it to leave the gate, taxi, takeoff, land, and return to the gate during the February demonstration?

Gariel: The technology we added can be explained in three main components.

An automated flight control system to control all flight control surfaces including the aileron, elevator, rudder, flaps, nose wheel, differential brakes, and flaps.

A Perception stack for obstacle detection and avoidance both airborne and in the air (Detect and Avoid (DAA). The stack consists of radars, ADS-B, cameras, and lidars. The DAA system is coupled with the flight control system to automatically avoid threats.

A redundant command and control link as well as telemetry link for flight test purposes.


Avionics: What type of ground control station was used to fly the Caravan, and how was the remote pilot able to track the movements of the aircraft from takeoff to landing?

Gariel: We are operating the aircraft from our mission control center located in Concord, CA. The control center receives state data as well as video using our redundant data link.

Avionics: How long did the flight last and what were some of the biggest insights or learnings you concluded after completing this flight about how your autonomous software stack enables autonomous gate-to-gate flight?

Gariel: This gate-to-gate autonomous flight lasted just about 10 minutes, but we repeated it over 30 times. Depending on the weather conditions, in particular wind, the results were slightly different. We also performed this gate-to-gate test on 3 different runways, taxiing along many different paths on the ground. This allowed us to understand the effects of runway slope and length, headwind and crosswind effects on the taxiing and flight dynamics. We also learned a lot about ground operations and interactions with ATC, understanding what kind of requests we need to be able to accommodate.


Avionics: You were able to manage all air traffic control interactions from the ground. Was the aircraft equipped with transponders or other equipment that was providing its updated position to ATC throughout the flight or was that managed in another way?

Gariel: We operate in the mode-C vail, so the aircraft is equipped with a Mode-S transponder and ADS-B.

Avionics: Are there any key sources of surveillance, flight, or other aviation flight environment or air traffic data and information that were key to enabling this demonstration?

Gariel: Our aircraft are equipped with Xwing’s Detect and Avoid system that fuses ADS-B, onboard radar, and vision system. The ground control station also receives ADS-B from a ground-based receiver.

Another image captured from Xwing's autonomous flight demonstration in February.Xwing

Avionics: In terms of next steps, how will Xwing continue to develop this technology to the point that it is ready for safety-critical civil aviation authority airworthiness certification? Do you have a timeline as to when you expect your software stack to be ready for certification?

Gariel: We work hand-in-hand with the FAA to bring our technology to market. We are running two tracks simultaneously.

The first track focuses on certifying individual components of the system through STC. Our first STC focuses on the detect and avoid system and is currently underway. STCs for the flight control system and coupling with the DAA will follow.

The second track focuses on optionally piloted and unmanned flights, which uses non-certified technology, but operational limitations to mitigate risk. This will allow us to perform revenue operations without requiring full certification. The operations will be limited to pre-approved routes initially until it is approved for further expansion.

previousFuture Military AI/ML: Multi-Plane Collaboration, Precision Targeting, Full AutonomynextAre Artificial Intelligence and Machine Learning Essential to the Success of the Commercial UAS Industry?