Applying intelligence to autonomous flight

Deep Dive

“Instead of thinking ‘AI’ represents some looming black box magic, we focus on what we can do with it here and now,” says Dr. Yemaya Bordain, Daedalean’s president of Americas. “AI is a loaded term these days.” 

The definition of ‘artificial intelligence’ basically means leveraging computing in new and more powerful ways. Daedalean wants to apply that intelligence to solve real aviation problems. Since founding in 2016, it has raised $72.5m across three funding rounds, including a $58m round closed last year. Lead investors include Carthona Capital, Honeywell Venture Capital and Redalpine Capital. The firm has also been awarded multiple grants including from the EU’s research and innovation funding programme ‘Horizon2020’ and, most recently, from the Swiss Innovation Agency ‘InnoSuisse’.

Daedalean is not reinventing the wheel, or the aircraft in this case, Bordain tells Revolution.Aero. The Swiss company is designing systems which make use of the recent expansion of computing power to increase flight safety and improve economics. “We want to be the key enabler of safe, efficient, accessible flight starting with pilot assistance and setting a trajectory for full autonomy,” she says. Daedalean is focusing on the “here and now” by taking the same starting point as pilots have since the beginning of flight: where they are, where they can fly and where they can land.

“Our systems allow smaller aircraft to benefit from the high levels of safety and efficiency that airliners enjoy but which, for airliners, require huge investments in ground infrastructure. Our technology can be implemented in nearly any aircraft, from private four-seater fixed-wings and commercial helicopters to AAM, whether eVTOL, STOL, or eCTOL, and commercial aircraft,” says Bordain.

Daedelean takes its name from Daedalus, who according to Greek mythology was the first person to achieve heavier-than-air flight. Daedelean can also mean “difficult to comprehend due to complexity or intricacy”, so in that respect the company will myth bust their own name when they certify. The first product with Daedalean technology is due to obtain FAA certification with concurrent validation by EASA soon. When it does, it will be the first AI in the cockpit. The product, called PilotEye, from US-based avionics firm Avidyne, will provide traffic detection functionality for general aviation. Through Daedelean’s technology, the PilotEye system will provide alerts for both cooperative traffic carrying transponders and non-cooperative traffic.

“We have fielded interest from operators and OEMs focused on HEMS, search & rescue, law enforcement, firefighting, and others,” says Bordain. “We’ve also talked to flight schools who see improved learning efficiencies in terms of on-stick instruction and analytic feedback applications. But the first use case could very well be remotely-piloted cargo operations, which would be able to fly BVLOS as the onboard system would serve as a backup when the remote command is too slow or suffers a lapse in connectivity. Not too far down the road, we imagine our technology could aid soon in time-critical trans-oceanic cargo transport.”

AI-integrated products are one thing, but a fully-automated pilot system is another. Bordain predicts full autonomy will be possible by the end of the decade, although crystal balls are never foolproof. “But it’s anyone’s guess as to when it will be widely available as a certified technology allowed in operation on civilian aircraft. However, our guess is that it will likely be mid-next decade.”

Overall, the system is platform agnostic and is viable on nearly any aircraft, Daedalean claims. The main difference between a fixed-wing and VTOL setup is the arrangement of the cameras. On a fixed-wing, it’s convenient to place the sideways- and forward-looking cameras on the wings and the downward-looking, fish-eye camera on the bottom surface, under the fuselage. For VTOLs, this is not possible, so we place all the cameras in a custom enclosure mounted at the front of the helicopter, with each camera looking either sideways, ahead or down. The software application set is different too. For a fixed-wing, it includes a runway finder and landing guidance. For a VTOL, the software has an emergency landing spot finder and vertical landing guidance. “Both landing functions operate independently of GPS and ground-based infrastructure, and across a wide altitude band–operating more precisely as altitude decreases, making them extremely valuable,” explains Bordain. The vision-based positioning system’s ability to operate independently of GPS, yet remain as accurate, is increasing its value, because threats to GPS have shown an uptick in recent years. However, Daedalean’s positioning system does continuously track GPS signals in order to merge the aircraft track defined from visual information with the GPS track. “This enhances the system’s ability to smoothly sustain navigation in case of a sudden GPS outage. As such, our visual positioning system is a strong candidate as backup, complementary, or alternative positioning and navigation,” says Bordain.

The first rule of a successful autonomous pilot system is that it doesn’t miss a hazard. Daedalean’s software can detect and categorise hazards using both cameras and sensors. The real-time video stream feeds into an onboard computer containing the firm’s machine-learned algorithms. It analyses the images to determine if there is an object in any of the camera’s twelve million pixels and then categorises any objects by aircraft type – helicopter, fixed wing, or something else. It can then determine size and distance and by tracking the object over several images, it can determine speed and direction.

“Future versions of the system will benefit from additional sensors such as LWIR cameras, radars, and possibly LIDARs. There is also the possibility of integrating our system’s traffic data with another DAA [Detect and Avoid] system on the aircraft,” says Bordain.

As for avoidance, there are two ways to go. The system can alert the pilot, who can then execute an avoidance manoeuvre. Or, by integrating the system with an autopilot, the autopilot can act upon the information provided by the system. Daedalean is currently in talks with potential partners about jointly testing an autopilot integration. “Risk level assessment is on our roadmap for development and will be key to creating what we call situational intelligence, which is the ability to evaluate, anticipate and act upon potential threats. This capability will take into account ownship velocity and manoeuvrability. But long before situational intelligence is fully developed, our camera-based system’s ability to detect and track the surrounding traffic just by looking around already has all the critical information about hazards: distance, direction, velocity, and size,” says Bordain.

Subscribe to our free newsletter

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


Leave a Reply

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