Drop the pilot or ‘keep skin in the game?’
What’s the fastest and most efficient way to develop autonomous aircraft – to drop the pilot from the outset or keep one on board, at least initially, to keep “skin in the game”.
The developments in the fields of artificial intelligence (AI) and autonomy make a compelling argument for fully autonomous aircraft. However, those in favour of remote piloted aircraft (and pilots) will say the human is best placed to deal with unforeseen events. Both groups agree regulatory bodies are not yet ready to certify autonomous aircraft in an urban setting.
Marc Ausman, co-founder and CEO of Airflow Aero compared the debate to one in the automotive industry – between self-driving car company Waymo and Tesla. “Waymo went directly to autonomy and they are doing limited trials in some cities, whereas Tesla, which put the driver in the car first, is now a $400bn business,” Ausman told Revolution.Aero’s Town Hall online – The Autonomous Aviation Revolution.
“It can use the tons of data collected along the way and this experience to mature its self-driving car.”
Airflow Aero is taking the Tesla route, putting a piloted aircraft in the field, which will collect data to eventually fly autonomously. “We will move to an autonomous platform when the regulation supports that,” said Ausman. He believes that autonomy is gated by regulations and not by technology.
“There are lots of companies building the autonomous technology that others can use.” The big question is how companies choose to get there. Ausman said: “Everyone wants to ultimately get to an autonomous aircraft, to remove the pilot and get the benefit of higher payload and cruise scheduling and potentially some cost savings down the road.”
Daedalean AI’s founder and CEO, Luuk van Dijk on the other hand, is unconvinced by remote piloting: “If you lose the commander control, you are flying blind. So, to deal with that, you need some form of autonomy. And if that is good enough, you can complete the mission as planned without humans. As a human passenger, I would hesitate to get in an aircraft with a pilot on the ground who does not have skin in the game.”
Van Dijk’s company is building the autonomous technology to eventually fly aircraft. Seeing how reliant today’s commercial aircraft are on human decisions, he said Daedalean is building a “machine equivalent of human capability, the ultimate form of autonomy which does not require remote piloting”.
To meet this requirement, it means giving the computer eyes, said van Dijk.
“Currently, uncertainty [with AI applications] in avionics is dealt with by passing it to the human. Now if we want autonomy, we have to deal with the uncertainty in the environment. We can use a harder class of systems to get to a higher level of safety, provided we do it right.”
Van Dijk says the next step is making sure regulators such as EASA and the FAA can see that these technologies are safe to use. Admittedly, machine learned systems are not perfect, but they are analysable and manageable. Autonomy will also mean more expensive systems on the aircraft to facilitate the human functions.
It is difficult to argue with the numbers, which show a high percentage – some as high as 80% – of aviation-related accidents are due to human error. With this in mind, Wisk CEO Gary Gysin’s statements about self-flying aircraft ring true. “It’s not that the humans do not know what they are doing, but the systems are incredibly complex. We have had more than 1,400 autonomous test flights with our vehicles without an incident,” he said.
But there are other factors to consider once the technology is mastered and regulated. AAM players will still need to gain public acceptance. “Not everybody jumped in a Tesla and turned on autopilot and just started going,” said Gysin.
Paul Smith, head of test flight and operations, Acubed by Airbus, said: “Both can learn from each other. If you map enough signals and get enough repetition, then you can use machine learning to start replicating decisions that pilots make.”
As with many things, finding a path of moderation might be the key here. For early trials in the short term, many including Smith have suggested the combination of the pilot and AI. And in time, perhaps this will lead to increased regulatory and public confidence, laying down an aerial path for a fully autonomous future.
Marc Ausman, co-founder and CEO, Airflow Aero | email@example.com
Luuk van Dijk, founder and CEO, Daedalean AI | firstname.lastname@example.org
Gary Gysin, CEO, Wisk | email@example.com
Paul Smith, head of test flight and operations, Acubed by Airbus | firstname.lastname@example.org