Understanding the shifting relationship between hardware and software is fundamental to the future of the digitised aviation industry, and to the technology stacks on which it is built. But how is the industry navigating those changes and challenges? We talk tech with key researchers looking into how the digital landscape is changing — and needs to adapt.
Felix Berteloot, deputy head of automation at the ZAL Center of Applied Aeronautical Research in Hamburg — and resident AI expert there — tells us that flexibility is key. “As our main focus is on bespoke solutions for our industrial customers we seldom interact regularly with the same business interfaces. Of course, state-of-the-art tools like Git are used for example for code delivery and dissemination, but in principle, business interfaces depend on the actual application.”
This challenge of balancing customised implementation with commonality of provision is echoed throughout the industry, not least with the substantial legacy hardware and software issues — and sizeable technology debt — that are inherent in aviation’s technology stacks.
“However,” Berteloot notes, “we do look into larger scale business interfaces, such as the European Gaia-X initiative that might be beneficial for our operations.”
Gaia-X, proposed in 2019 and established as a non-profit in 2020, intends to establish a federated secure data infrastructure developed on European values and principles. Its membership includes wider industry like Siemens, Orange and EDF, as well as industry players like Safran, Deutsche Telekom (which partners with the former Inmarsat, now Viasat, in the European Aviation Network hybrid connectivity system seeing renewed traction recently) and Amadeus.
More widely at ZAL, Berteloot says, “we use a combination of both established and robust technologies as well as modern and novel technological approaches. In particular, these well-known and established technologies range from things like programmable logic controllers (PLCs) to robust programming languages such as Python and C/C++.”
On the hardware side, “examples for our novel technological approaches include autonomous ground vehicles (AGVs), industrial robots and cobots with various end effectors and methods of artificial intelligence, such as computer vision and deep reinforcement learning — to name a few,” Berteloot explains.
Current combinations of software, meanwhile, include a variety of programming languages, AI frameworks including PyTorch, containerisation techniques like Docker, and other technologies familiar within the wider tech sphere.
“We are always looking for further improvements in our tech stack,” Berteloot says. “Sometimes tech stack changes simply mean upgrades to newer versions such as the move from the Robot Operating System (ROS) to its successor ROS2. Sometimes new possibilities arise, for example with the launch of NVIDIA’s new Omniverse suite or similar tools from other companies that can be used for the generation of synthetic training data that feed our deep neural networks.”
As on-aircraft connectivity spreads more widely, it enables new opportunities for operators
From commercial airlines to private operators, the benefits of inflight connectivity are already changing the digital aviation landscape. Only a few low-cost carriers at this point eschew broadband satellite or hybrid satellite-passenger connectivity, while more private aircraft are connected than ever before.
On the business aviation side in particular, Craig Foster, senior consultant and cofounder of Valour Consultancy tells us, “we’re really at the intersection of a technological revolution that will see the hardware-centric approach of old make way for a future defined by software.”
For private aviation, the relatively smaller size of aircraft (compared with commercial airliners) has led to a new generation of bizjet-specific hardware enabling smaller airplanes to be connected just as well as — if not better than — their larger cousins.
“We’ve seen companies like Satcom Direct with the Plane Simple line of products seek to develop their own line of hardware, designed specifically for the requirements of business aviation. Previously, it was common to repurpose equipment from the commercial aviation realm,” Foster explains.
This raised a number of issues around larger pieces of connectivity kit like antennas and terminals, especially those for Ka- or Ku-band connectivity, which have tended to be both large and heavy. This proves complex in terms of physical space, aerodynamics, performance and other considerations.
“For antennas, specifically, the size of the airframe is the obvious limitation,” Foster says. “It’s impossible to fit bulky, mechanically steerable satellite antennas on anything smaller than a super-midsize jet, hence the excitement around new ESAs [electronically steerable antennas] with a much sleeker footprint that promise to significantly expand the addressable market for high-speed connectivity.”
At the same time, while space for connectivity hardware and other electronics hardware can normally be found in avionics bays, cargo holds or ceiling spaces in commercial jets, this is a tighter squeeze, often quite literally, on a business aircraft. As a result, hardware options that are ruggedised for installation outside of temperature- or pressure-controlled areas of the aircraft are finding a niche.
AI remains the big beast of looming technologies
In discussions across the industry, it is clear that the biggest question is how artificial intelligence will affect the way that aviation works, and that integrating it into the industry’s technology stacks will be crucial.
Despite questions of a hype bubble within and outside the industry, Berteloot acknowledges, “the hype goes hand in hand with a technical revolution as hardware, software and data availability are finally ready for real use cases. Although it will take a very long time — if ever — [that] an artificial intelligence steers an airplane, the technology will soon be used in virtually every technical domain, from production planning to manufacturing, from ground handling processes to cabin systems and entertainment.”
In many ways advanced algorithms — for example, those that can identify types of ground vehicle at an airport, or flag aircraft operating outside a series of norms to air traffic controllers — already might fall under some definitions of AI.
One key challenge is keeping up with the leading — but not bleeding — edge of technology.
“Especially in terms of artificial intelligence, employed software, frameworks and repositories change rapidly,” Berteloot concludes. “Since 2019, we gradually moved from, for example, TensorFlow to Keras to PyTorch. In principle, we are technology agnostic and are always looking for the best performing tools. A main focus of ours is the use of open-source technology and software. The reasons for this are obviously cost, interoperability with other systems, robustness and the possibility of customisation.”
Here especially, the balancing act between developing that kind of customisation for solutions and ensuring commonality when it comes to supporting and servicing them will be vital. It’s clear that defining and navigating the relationship between hardware and software, together with the opportunities and challenges of integration within the industry’s technology stacks, will be critical to aviation’s digitalised future.
;
