DASC 2022: Past, Present and Future of Aviation Safety

On September 20, 2022 I was fortunate to have the opportunity to give the opening keynote at the IEEE/AIAA Digital Avionics Systems Conference in Portsmouth, VA. Below is the raw text of my talk. If you prefer to watch the video, it's on YouTube.

I attended the 2018 DASC in London and was really struck by what I experienced. Conferences I had attended up until that point were either too academic, focusing on abstract futuristic concepts with little consideration for how something would be implemented in the “real world,” or were too commercial, trying to sell you yesterday’s technology. DASC occupies this really interesting space in between, focusing on what we can do today and in the very near future, with an intense focus on pragmatism and safety that I think is really healthy.

Reviewing the program I’m really excited by the agenda this week. I wish I had more hours in the day to read all of these exciting papers. There is one session that jumped out at me, and I don’t mean to pick favorites or anything, but I think this title perfectly captures this spirit of pragmatism and safety and what excites me about this conference. It’s called “Enough Fluff: Returning to Meaningful Perspectives on Automation.” I hope it lives up to the title.

Reliable Robotics

We started Reliable Robotics out of a shared vision for making aviation safer and more accessible through increased automation. At the time, and perhaps this is still true today, we saw a gap in the industry between what was currently available, yesterday’s technology and the futuristic concepts that were being pursued by others. We thought there was space to find a middle path, to take what already existed, add a little more incremental innovation on top, and create something new.

As you may know, our first project is to certify an advanced autopilot for the Cessna Caravan, a popular fixed wing aircraft capable of carrying about 3000 pounds, or up to 13 passengers in some configurations. We choose the Caravan for three reasons.

One, we wanted our technology to have an appreciable impact on safety from day one. As any Caravan pilot will tell you, they do “real flying.” These aircraft operate in more challenging conditions than most other commercial aviation aircraft, at lower altitudes, in more serious weather conditions, and they do so single pilot. I have tremendous respect for Caravan pilots. These are true professionals doing one of the more difficult jobs today in aviation.

We felt that an advanced autopilot, one capable of running these operations end-to-end, would bring immediate safety benefits to these operations in terms of alleviating workload, improving flight planning, and less overall risk when operating in lower visibility conditions.?

Second, we wanted our first product to have a public benefit as well, and the Caravan stuck out for us because of the essential role it plays in providing access to rural communities and harder to reach corners of the planet. The town where I grew up, for example if you wanted to overnight a letter or receive an overnight package, the only way it was going to get to you was on a Caravan. I don’t think many people realize the essential role small aircraft play in connecting rural communities to major population centers, and the even greater role they could play if we could improve the safety and efficiency of these operations. I may be a bit biased growing up in a smaller town, but when I think about the promise of aviation and the benefits it can provide to society, I believe we should be thinking about the benefits it can provide to everyone, not just people that live in big cities.

Third, being a business, we obviously needed to start somewhere where there was a strong customer need. And the demand for automation of small aircraft is very high. And perhaps not for the reasons you may think. I get comments all the time about “costs” being a driver and that’s certainly part of the equation, but it’s not the number one reason air carriers want more automated aircraft. It’s about improved safety and increased flexibility.?

We choose an approach to automating the Caravan in a way that we feel strikes a balance between innovation and cautious pragmatism. Our system operates within the existing airspace framework and borrows heavily from existing accepted technologies, standards and procedures. I honestly don’t view what we’re doing as all that special, at least from a technical perspective, because I feel that we are merely pushing forward along this steady march of safety improvements that we’ve been at now in aviation for over a century.

Incremental Safety Improvements in Aviation

Innovation in other industries tends to be talked about as “revolutionary” or “moonshots” but aviation doesn’t work that way. What makes aviation special is its steady drumbeat of progress, continuously improving and evolving to make it safer. At Reliable we are just continuing this tradition.

Looking back over the last century, aviation has seen remarkable advancements over time. One of the first problems impacting aviation safety was navigation at night, and one of the first solutions was remarkably simple: light beacons. This concept was borrowed from sea navigation, but it worked and went a long way towards preventing nighttime accidents. One aspect of this that I thought was clever by the way was in some places the light would change color based on the viewing angle, so you could tell if you were off course. And if the light disappeared you were either way off course or there was a cloud–or mountain–in the way.

Light beacons were eventually replaced by radio navigation, initially NDB and then later VOR, as these were more robust and less susceptible to interference from clouds! And then Instrument Landing Systems were created to provide more accurate guidance getting the aircraft back to the airport environment.

Many of the concepts, technologies and procedures for ILS carried forward into autoland. Autoland systems were developed in earnest in the 50’s to address safety concerns with low-visibility operations in the UK and European winter. This early work spearheaded concepts of “target level of safety” and fault tree analysis, ideas and processes we use almost everywhere and in many different industries today. Autoland also drove some of the first work into triple redundant avionics, another safety and reliability enhancement that is used for many subsystems on aircraft today, not just autoland.

One of the first DASC papers describes a modified 737 capable of 4D path planning and flight control, and this vehicle was later part of the NASA’s Terminal Configured Vehicle program in the late 70’s. By the way they called this system the Advanced Electronic Display System or ADEDS, which is a terrible acronym for a safety enhancing system, or maybe this is how their engineers ensured prototype code would never be used in production! This system was capable of complete flight from takeoff through landing, and many of the techniques it describes builds on previous work that was done for the development of autopilots and autoland systems. And again, the motivation was to maintain and increase safety.

Terrain avoidance and traffic avoidance systems then built on this work, driven by a desire to further reduce flight into terrain and mid-air collisions, which were shockingly common in those days.

I’m oversimplifying a bit, obviously, but I’m trying to make the point here that when you look at what we have today, we now have an incredibly safe air transportation system because of this long steady stream of incremental safety improvements. Every safety innovation has built on the last. GNSS navigation, GPS, WAAS, these systems really build on what had existed previously. Sure, the underlying technology is very different but the methods and procedures for getting them into aircraft and the airspace you can trace back to earliest navigation systems.

Aviation Isn't "Too Slow"

You hear people make comments about the pace of innovation in aviation being too slow, and I just don’t see it that way. Sure, I’m an engineer, and I’m an entrepreneur, which makes me an impatient engineer! Things can always be done faster and better. But I don’t see aviation being slow for the reasons many people describe.

Aviation–civil aviation–is slower not because it takes time to adopt new ideas, but because in aviation you don’t make changes unless you can make a solid case that the changes make things safer. But it’s not a single criteria that decides what is safe or what is not, it’s a rich set of interconnected organizations, tools and processes that tie everything together. It’s like the Swiss Cheese Model, but on a macro scale, or you could think of it as a lot of interlocking gears. Aviation’s pragmatic focus on incremental safety improvements is not so much a mantra but more a result of all of these interlocking gears working together.

Future of Aviation Safety

We have many more incremental safety improvements right in front of us, many of which we are going to realize in the very near future. I think this is a really exciting time to be involved in aviation. I want to mention a few of these are really exciting to me, not just for their long-term potential but also the safety benefits they can bring to aviation in the near term.

Single-output navigation systems that are 10-8 or greater are right around the corner. This has been a big focus for us at Reliable, as this will be a core enabling technology for many things in the future. What I mean is a navigation device, or an ensemble of devices, that give all other aircraft systems position, orientation, velocity, etc. with sufficient robustness and redundancy that it can be trusted for all critical operations. The safety benefits of such a system will be tremendous. I think about it as a pilot, you know you’re trained to check and cross-check your instruments against one another? Don’t you want the system to do that for you? Just tell me what I can trust? Or better, just be one fault tolerant? Instead of telling me what I can and can’t use, tell me when it’s time to land because I’m one fault away from a bad day. That’s what you want.

4D terrain-aware flight planning is almost here. We are so close to realizing the dream of the Terminal Configured Vehicle program. We will very soon have systems that allow you to fully specify a 4D flight plan from takeoff to touchdown, in a certified way, while still accommodating real-time changes to that plan to meet dynamic traffic control and weather requirements. This matters a lot because controlled flight into terrain is still the number two cause of fatal accidents in General Aviation, and commercial airliners are not immune to this either. We have the technology. People don’t need to die.

Once you have 4D flight planning now you can build a system that continuously engages the autopilot through all phases of flight. What I mean is, you click the autopilot on at the hold short line and you don’t click it off until you’re off the runway at your destination. Building this type of autopilot is not hard, there’s no fundamental research problem to solve, it’s just millions and millions of little details. And this matters a lot too, because Loss of Control is the number one cause of accidents in General Aviation, and the commercial airliners are not immune to this either. This also matters for pilot training–simplified, flight interfaces will make knowledge transfer between aircraft smoother, and thus safer.

Autoland without infrastructure at the airport is another very exciting area. There are a lot of proposals out there right now, some I saw in this DASC program, that use all varieties of new sensors, or combine sensors with existing navigation sources in new ways. Being able to perform a CAT III C autoland on more aircraft to more airports is going to make a huge difference when it comes to safety and availability of aircraft operations. For commercial ops especially, I’m excited by the number of airports that this will open up and the massive number of delays due to weather that will be prevented.

Conversely, auto-takeoff is an area where we are extremely close. The takeoff actually isn’t the hard part–most aircraft “want” to fly. The tricky part is takeoff rejection. This is an area where automation can really shine. Engine and other critical system parameters can all be monitored hundreds of times per second and make a decision much faster than a human. I was just watching a video recently where a GA pilot was analyzing a rejected takeoff and he admitted it took him a whole four to five seconds to realize his engine pressure was in the yellow. He was fortunate he caught it before taking off, but he ate up hundreds of feet of precious runway in the process. And similar to CFIT and LOC, aircraft misconfiguration on takeoff is also a common cause of accidents, and auto takeoff will go a long way towards preventing these types of accidents.?

In the near future we will also have autotaxi. Many of the building blocks are almost ready for use in the current airport environment. Navigation databases with taxiway data exist, high-precision navigation systems are almost ready. You take these and combine them with the same sorts of components and control techniques for auto-braking following landing and you’re almost there. And insurance companies are going to love this!

Certified, bi-directional communications for command and control is very close as well. We have many systems for communicating with aircraft today, but we don’t yet have anything that allows an operator to remotely command an aircraft, and remotely command an aircraft in a way that would alter its flight path. CPDLC is close but this is designed for “non urgent” communications. This type of system would make air traffic control communications much more robust and less open to misinterpretation. This would also make communications more secure and allows us to begin to layer new more sophisticated features on top.

Detect And Avoid (DAA) is another key area. The first challenge here was defining what the D and the A actually mean. RTCA SC-228 and countless research teams spent years going back-and forth on how to define aircraft separation, or what it means to remain “well clear”. Going back to one of my earlier points, this is partly why things in aviation take as long as they do. In order to solve a problem, and do so in a way that maintains or increases safety, you first need to meticulously define the problem you’re actually trying to solve. You don’t just jump in with a solution.

It’s exciting to think about what you can do when you start to combine all of these components I just described. Imagine an aircraft with 4D path planning, autoland, auto takeoff, autotaxi, advanced communications, and new sensors and avoidance methods to maintain well clear. Now you can remotely pilot this aircraft! Think about how much safer this is going to be: If you’re operating the aircraft from a control center, you can have people cross-checking decisions. You can take Crew Resource Management (CRM) concepts and scale that up to a team of people. You can borrow techniques from Air Traffic Control management. You can have caution and warning systems that alert other pilots. Crew rest and fatigue considerations can be managed in an entirely new way. You could even have shift changes in the middle of the flight. Remote piloting is going to drive a lot of improvements to aviation safety, some of which we probably can’t even imagine today.

I also want to touch on Digital Flight. This idea has flown under a lot of different banners, some call it UTM, Free Flight, or Advanced Air Traffic Management, but the core idea is the creation of new operational rules that use the latest advancements in technology to rethink the airspace, and drive higher levels of safety and efficiency, much how the airspace was redesigned for the jet age. I feel this is incredibly important work and much more work will be required if we are to realize this vision. We especially need to consider how we’re going to integrate existing traffic operating under VFR and IFR with remotely piloted aircraft–at scale. To my point earlier, aviation changes are driven by safety imperatives, so it's on us, this community, to show that this new system will be safer.

I would be remiss not to bring up security as well. Combining all of these components is going to raise legitimate concerns about the security of these operations, potential vulnerabilities and the potential for misuse. There are many building blocks to draw on from aviation and other industries but every new use case introduces new threat vectors that will need to be considered, and this community will need to face them if we are to obtain the public’s confidence in these systems.

A Great Transformation

I believe we are on the cusp of a great transformation. And I will be so bold as to compare it to the Internet! People talk about the Internet as some kind of revolution, I think you’re picking up my theme here–it really wasn’t. The Internet was the culmination of almost a century of incremental innovations in electronics, transistors, algorithms, programming languages, and communication systems. When I was a teenager, in 1992 O’Reilly published the Whole Internet User’s Guide and Catalog, if anyone remembers this. They compared the “world wide web” (what we now think of as the Internet) to Gopher, which was really just a fresh take on FTP. At the time the web, HTTP and HTML, really didn’t seem like anything all that special, it was just an incremental update on prior technologies. I think very few saw how this new web thing was going to take off and radically change the way we live and work and communicate with one another.

I think we’re at a similar stage right now with aviation. Looking back, countless numbers of small incremental improvements have made aviation safer. But these incremental safety enhancements are about to be combined in new ways that’s going to make aviation, like the web, radically more accessible to more people.

Conclusion

I hope you found my perspectives on automation meaningful, and not too fluffy. At the risk of introducing a little bit of fluff, I do want to stress how important the work this community is doing and how much we value it at Reliable. Fully automated aircraft are going to be aviation’s “Internet moment”. Think about all of the additional air traffic we will be able to drive into the air, the efficiency gains and the improvements to affordability that we’ll be able to get. Think about all of the additional people that will be able to utilize the air for work and travel, and how much smaller this is going to make the world feel. Think about the lives that will be touched, the experiences that will be had, and the families and loved ones that will be connected.

I’m sorry for getting fluffy, but aviation is just such a beautiful invention, and the collaborative nature and focus on safety is equally beautiful. I’m proud to be a part of it. On behalf of Reliable Robotics, thank you to all you for the work that you do, and thank you for having me here today.