Stacking perception: Xwing and NASA combine on autonomous flight

Deep Dive

The largest elephant in advanced air mobility’s room is autonomy.

Making an aircraft that can safely take off, fly and land by itself is one thing. But building an aircraft that can perceive and adapt to constantly changing environments is a different challenge altogether – especially in an urban setting where there are infinitely more hazards and lives at risk.

Autonomous flight processes need to be developed to allow these aircraft into domestic airspace, without them there is no path to certification.

Xwing has just landed a three-year contract with NASA to share data and build safety procedures for both autonomous operations and design. Led by NASA’s System-Wide Safety Project (SWS), the contract firstly aims to evaluate safety arguments around runway detection and identification for vision-based landing. It will also assess aircraft localisation assurance processes and enhance GPS, according to NASA.

“There are different components associated with doing this,” says Marc Piette, CEO, Xwing. “A big part is the safety case, that comes with system safety analysis, data analysis. To ensure these vehicles can integrate safely with other traffic whilst being able to perform all the phases of a set mission.” 

NASA uses Space Act Agreements as a way of openly collaborating with various commercial entities across a range of fields. The agency currently has other agreements with Joby, Overair and Raytheon. Also Xwing will likely continue to work with NASA beyond the three-year contract.

Jesse Kallman, vice president, Commercialization & Strategy, Xwing says: “We are now openly able to share our own flight data, both from our autonomous system and Part 135 operations. There’s tons of information to be gathered from Part 135 ops.”

Xwing recently expanded its unmodified fleet and now operates 400-plus weekly cargo flights for UPS.

“On the flip side, NASA has also been developing a number of new algorithms around things like automated landings and integrating with airport patterns. All of that information is super-helpful for us as we look at how we build a better safety case to bring to the FAA to get these systems certified,” explains Kallman.

Automation has been around for some time. Maxime Gabriel, chief technology officer, Xwing was the chief engineer on a successful fully-automated helicopter programme at Collins Aerospace. The helicopter could even land on the back of a moving truck.

“It had the safety level of an airliner,” says Piette. “For a catastrophic failure, it was very close to 10−9 per flight hour there. So that has existed and people have done this. The main challenge here is how do you bring a highly-automated vehicle and integrate it safely with ground and air operations, with air traffic management.”

(The parameters for airliner safety: “Extremely improbable” is defined as 10−9 per flight hour, or one in 1bn flight hours)

The helicopter was fully-automated but it was not autonomous. Put a wall in front and the helicopter will have no qualms flying into it.

“It doesn’t have a perception stack. It doesn’t have a very good understanding of its surroundings,” adds Piette.

The system needs failure modes. “There’s technology out there that has great capabilities but reduces safety levels by assuming the pilot is incapacitated,” says Piette. “It relies on technologies like GPS to bring the aircraft to the ground. The problem is GPS integrity levels are too low to be used for nominal autonomous operations.”

How to enhance GPS without the need for expensive ground-based infrastructure like a category III instrument landing system (ILS)? (CAT III approaches use radio altimeter (RA) to determine decision height).

Xwing is using a camera-based system to automatically identify where the runways are. “Based on where that runway is on the camera feed, we can use our algorithms to map out the position of the aircraft in space. This gives us a totally independent way of positioning the aircraft.

“That is one of the capabilities that this data collection is used for,” says Piette.

American writer Walter Lipmann once said: “You cannot endow even the best machine with initiative; the jolliest steam-roller will not plant flowers.” He will soon be proved wrong if all goes to plan.

Subscribe to our free newsletter

For more deep dives from SAF Investor, subscribe to our email newsletter.

Core topics

Leave a Reply

Your email address will not be published. Required fields are marked *