The Flight Automation Era, with Mark Groden (CEO of Skyryse)

The technology in these cockpits has not gotten simpler over time. It's gotten more complex because nobody's done a clean-sheet approach to how do we redefine the human-machine interface as we have to make it safer and easier. The fact that we could get you into a helicopter and in 15 minutes you were literally in full control from pickup to set down to flying around the greater Los Angeles area. There's no vehicle that you can have that level of empowerment with that little training. And that's not an accident.

History has taught us if you make transportation a little bit easier, a little bit more accessible, people don't consume just a little more of it, they consume dramatically more and then the infrastructure catches up.

Thanks, Mo. Great to be here.

Happy to be here. Yeah, funny status.

Yeah, I guess so. Just the high level of what happened there, it was a low-altitude stall in a glider. And basically the center of gravity was too far aft. There's a tail wheel that was left on the airplane at takeoff. There's like these very large kind of red tail wheel casters that go around the back of an airplane like this to maneuver it on the ground. And it's large and red so that you remember to remove it before you take off. Unfortunately, this particular day it wasn't removed. And so the tow line broke that was pulling this — gliders aren't powered — pulling the glider up, and as it did, center of gravity was far aft, stall, which basically meant that they pitched up and then went straight down into the ground. And so that was, you know, obviously tragic and terrible for three generations.

What I took away from that was, number one, you can be a really good pilot and still make a mistake, and so it doesn't take much to find yourself in harm's way, and oftentimes a single mistake can be unrecoverable, as was the case there. The next thing was, of course, being kind of a budding engineer and pilot, I'm opening up the statistics books and I'm going, well, how often does this really happen? And to my dismay, I figured out, well, this is like every couple days in the United States, a pilot makes a mistake and then it's an outcome that's as tragic as that one. So that was really difficult to kind of wrap my head around. And that and a few other things that, you know, touched me over the course of my life growing up really pushed me in this direction. And when I was finishing my PhD, I looked around the industry and I didn't see anyone that was directly working to solve these problems. And so 10 years ago I founded Skyrise to make it, you know, safer, easier, and make airplanes, helicopters that much more mission effective.

Correct.

So specifically it's in Bayesian networks, which if you know Bayes' theorem, which is like probability of B given A. What's the probability that it's going to rain today if it's cloudy and whatever other factors you want to include. I was really fascinated by this field because it's fully deterministic, meaning you can take all the intelligence, the statistics of what's going on in the atmosphere, and you can encode it into a Bayesian network and then say, okay, given some particular thing that's observed, like there's a cloud today, and whatever other factors you want to include, what's the probability of some given outcome? And then you can layer in expected value math into that for what we call utility functions, which are like what is the value of any given outcome, and then use that as a basis for intelligent decision-making.

Not directly, and I mean it's important to go steps before that. Like, I was building radio-controlled airplanes when I was a kid. I was working on small drones as a part of an Air Force research project when I was 15, 16. Built a vertical flight capable, small, backpackable, fixed-wing aircraft. Worked on problems relating to landing drones on aircraft carriers. And in the course of my undergraduate research, I came across this kind of area in Bayesian networks. And something just really struck me about using this type of math to support decision-making. And so at the time I was applying it to a lot of different things. One of which was like ships, if they take a hit, what do they do? Do they go to shore? Do they try to, you know, stay out at sea and evade the enemy? How do you real-time make intelligent decisions about a degrading vehicle?

And the irony is it's hard to say how much of that expertise, quote unquote, that I developed during my PhD is actually applied to our product. But I actually think the frameworks that I use for decision-making have been more useful to me as a CEO, because as a CEO you have to make a lot of decisions with very limited information, and if you wait until you have perfect clarity of the picture, you probably waited too long to make the decision. So decision-making in high uncertainty, I think, is a huge part of my job, and I basically studied that, you know, for my entire pre-PhD.

Yeah, the core belief early on was that existing airplanes, helicopters are not as safe, not as mission capable, and not as easy to fly as they could be if you applied today's technology to those aircraft. And they're actually pretty good at the flying part. So that is converting some stored energy into motion through air and delivering people or cargo to the other side of a mission. You know, the airfoil, the engines that are used, that's pretty refined technology that's had, you know, hundreds of billions if not trillions of dollars invested in over the century. The shortcoming has really been the user interface and the digitization. So if you look at airplanes and helicopters, until Skyrise, they are all hardware-defined and software-enabled. And that's the big kind of pivot or change that we're ushering in. We're saying no longer should they be hardware-defined, software-enabled. They should be software-defined and hardware-enabled.

And in so doing, you can make those aircraft that much more mission capable, easy to fly, safe. You can go from two pilots to one pilot to no pilots if the mission requires it. And it gives you so much more flexibility in terms of whatever features and mission-specific capability sets you want to be able to offer. The other thing that I would point out is the development of a whole new airplane or helicopter is incredibly complex. And I think this is one of the things that I learned in the course of working on a drone research project when I was a teenager. I realized how hard it was to build a small backpackable drone. And that was, you know, there's no humans on board that machine. There's no FAA to talk to. It's just the design and engineering of an entire flight vehicle platform. And that is incredibly challenging to do.

Now when you put humans in the mix and then obviously you put the FAA process on top of that and you have to undergo a whole type certification process. I mean, this is layers and layers of complexity that you're adding in the development of a machine. And the other element is if you are to develop a machine, you are investing in a very specific set of capabilities — speed, range, payload — and you can't change those, you know, as you're developing it and learning things and eventually fielding it in the market. So all that to say, I believe that the bigger opportunity was as it relates to improving upon the existing fleet of airplanes and helicopters, both on a retrofit basis because there's 400,000 of them in the field right now, and then obviously for people building new aircraft, wouldn't it be great to build them without all that mechanical stuff built in and instead build them digitally at outset and offer that much better performance?

Yeah. No, appreciate you recounting the story, Mo. I guess before I answer your question, just a reflection on what you shared. We're really proud of the degree to which we've been able to simplify the flight controls of the aircraft. The fact that we could get you into a helicopter and in 15 minutes you were literally in full control from pickup to set down to flying around the greater Los Angeles area. We always say there's no car you can get into after 15 minutes of training. Park, go on the highway. Parallel park, certainly not happening, right? There's no vehicle that you can have that level of empowerment with that little training. And that's not an accident. That's again because the vehicle became software-defined years before you got into it with our technology, and then we designed and tested different versions of the human-machine interface for literally years on a daily basis until it got to the point where it was as you experienced it. And we've collapsed that learning curve at least for the basic aviation element of flying the aircraft. So we submit that we made the helicopter easier to fly than driving a car, which is a huge statement. And so, you know, you can imagine a future where anybody who can drive a Humvee can hop into a helicopter and in a pinch, you know, fly themselves to safety. So we're really proud of that. And that's the first product that we're shipping.

To your question about what we've integrated the technology into, so we started with the hardest technology problem first, which is helicopters. Helicopters are inherently unstable. So what does that mean? A helicopter, if you let go of the controls, it will exit what is safe and spiral out of control and become unrecoverable. That is not true in an airplane, nor is that true in any other vehicle on the planet that we know of. That means from a flight control standpoint it's that much harder. Weight, volume, cost — because this is a trainer helicopter, it's a very low-cost machine — those are all very hard requirements that we imposed on ourselves by taking that machine as our first one. So we started there. The good news with starting with the hardest is once the technology is done and applied successfully to that aircraft, it's all downhill.

So if you go to the bookend, I would say the other end of the small helicopter in the helicopter continuum is a Blackhawk. So we got a Blackhawk at the end of last year, and in 91 days from the first wrench touching it to flying fully automated with Sky OS, and you could hop in, Mo, and do exactly the same thing that you did in the R44. That was 91 days.

So we're really proud of that, right? Because we're able to take what is now COTS technology off the shelf, plug it into an aircraft as big and as complex as a Blackhawk, which believe it or not is actually easier because all of the requirements that I just walked through are that much more relaxed in the case of the Blackhawk, which, you know, from an outsider it's like, how could the Blackhawk be easier than a small aircraft? So that's kind of the bookending of the helicopter market. And then to your point, yes, we've applied that technology also to an SR22, which is arguably the hardest in the airplane category because it's the smallest single-engine trainer, generally speaking, airplane. It is the best-selling, which is part of why we started with it. But it is dramatically easier than the helicopter because the airplane, as discussed, is actually more stable than the helicopter, right?

So very easy application of technology.

Thanks, Mo. Yeah, we're really excited because we've been working with the defense market for the better part of a year, and so at this point we're actually formalizing that into Skyrise Defense, where we're deploying the SkyOS technology to defense-focused applications. And so that's crewed, that's unmanned, that's fully autonomous aircraft. It's basically one product with multiple aircraft, multiple missions, and now, you know, two markets. We've already proven the ability to apply it to the Blackhawk, but we've gotten so much more interest from defense markets in specific, even autonomy applications. Now we've decided to formalize that in a business unit, Skyrise Defense, and we'll have a number of product announcements coming in the next few months.

For the time being, all we can share is the Blackhawk. I would say the significant pull from this market has been, you know, can you take the pilots out of the aircraft, you know, for some specific missions? And the answer is absolutely. We believe we have the only optionally piloted technology in the world, meaning you and I can hop in the helicopter, airplane, fly it the way that we've been talking about, and then also hop out of that aircraft and it can fly in a fully autonomous, unmanned capacity, which is really powerful because right now these assets are being used in one way or the other, and you can imagine the aircraft flying to you, picking you up, and then you flying on.

Whatever level of control you want.

Yeah, so it goes back to the fact that we started with the hardest platform first, and by starting in the civil space with a light utility trainer helicopter, we had to design and engineer an off-the-shelf solution that's also FAA certifiable and achieves a price target that makes its way into an owner-flown aircraft. Meaning base MSRP of a Robinson R66 is about a million dollars. Equipped with our technology, there's about a $300,000 price increase from those as normally equipped, which goes from 1.7 for normal equipment and then 2 million for SkyOS equipped.

And when you're talking with the military, you know, the concept of taking what's roughly a $60 million Blackhawk and being able to apply a technology like SkyOS that's, you know, single-digit millions and offering this level of performance, set of capabilities like that's never before been thought about. The idea of having a technology stack that can be applied to a Blackhawk, increase the mission effectiveness, fly in zero visibility, you know, swipe down with your finger and land, and then fly that very same Blackhawk in an unmanned configuration. And oh, by the way, the technology that's enabling it, that's also, you know, about to be FAA certified. So we know the level of reliability is 10 to the minus ninth, at least as it's been engineered to.

That gives people a lot of confidence as it relates to some of the people who have been developing technology that looks, you know, similar for long periods of time. I'd say, you know, north of a billion dollars been invested in the development of these technologies. As far as we know, it's never been deployed and not been purchased, and it really wasn't designed as a product. It was designed to be a demonstration of certain technical capabilities. So again, getting to that cost point, you know, not FAA certifiable, not at a reasonable cost, and I'd say if it worked, it probably would have already been bought many times over.

Yeah, it's our fundamental belief that if there's a human in the back of the aircraft, one of those humans is going to be in control of it because the human plus the automation makes for a safer, better outcome. Why is that? Well, especially as it relates to defense missions, you have complex, evolving circumstances, and the human in the cockpit is aware of what's going on around the machine and is in direct communication with all the other aircraft around them. So the ability of that human to make real-time decisions about where they're going and why they're going there and adjust — you know, maybe there's an enemy shooting at them — we think is really important.

If the mission doesn't call for the movement of people, then that really begs the question, why put the people in the front and put them at risk? And so this is like logistics and ISR, and, you know, there's a lot of other missions that don't involve moving people. So if you can save the crew time and reserve those pilots for moving people and doing the, you know, the special missions-type operations, then that makes a lot of sense. And today the logistics burden on the military, I mean, it's massive. Like aviation supply is not there. The demand far outstrips the supply there. There simply are not enough pilots, not enough helicopters.

And right now, you're consuming the pilot time, right? The same pilot that could be flying, you know, people to the starting point of a mission are currently being used to run beans and bullets because it's the same aircraft. And if you can have the aircraft flying, you know, through the night doing those logistics missions as an optionally piloted vehicle, that completely changes the available time for the assets to do the more complex things. So automation actually gives you significant force amplification opportunity without even needing to buy new assets.

Yeah, it's hard to isolate any, you know, specific element, say, hey, this is 30%, 40%, etc. But it really kind of all comes together because at a fundamental level what we realized was to deliver the value propositions around mission capability and safety and ease of flight, right? Ease of flight is actually necessary to drive the safety. So today when people talk about pilot error, is it really the pilot making a mistake and being negligent? Almost never. It's the pilot getting inundated with tasks, and, you know, as soon as something goes wrong they're managing that in addition to everything else they already were, and then you have a cascading set of problems.

So when you put the full scope of the problem statement on the board, what you quickly realize is you have to control literally everything on board the aircraft through the system. And that means again taking the system from being hardware-defined to software-defined. So the operating system and the lines of code control or are part of controlling every system on board the machine. So take for example the Robinson R66. Inside that helicopter, the only thing that is not our bill of material on a systems level is a Rolls-Royce engine. That is a huge statement, right? So when you think about that, obviously when I started the company, that wasn't the aspiration. The aspiration was how do we deliver all this value proposition without having to re-engineer literally everything inside the airframe. And it was because of this conclusion we ended up having to take on everything.

And that includes redundancy and dissimilarity of different sensor types and the human-machine interface that you touched on and everything else. And so at the end of the day when you're touching all of the systems, it does give you the opportunity to completely redefine how you fly. And it turns out it's actually harder to certify that than it is to certify a new airframe in some respects because the complexity of a system managing everything and embodying the expertise of literally generations of pilots into an operating system while retaining 100% agency with the pilot is incredibly complex. Right now we've solved that problem — that, you know, it can only say in the last year or two — but now it's done and it does all those things.

So full flight control, right? Whether the aircraft is flying intact and everything's functioning. Flight control inclusive of emergencies, that goes as far as even if the engine is no longer on or working. Sensor fusion so you know where you are in space even without the GPS system. And then safety. Safety normally today in a cockpit, if something goes wrong, you pull out the pilot's operating handbook and you're literally going down a checklist. That's what Sully Sullenberger did in that, you know, the Hudson River scenario. For helicopter pilots, you don't have a spare hand because both hands are on the control. So you have to memorize the entire thing. Good luck remembering any one of the 300 pages. You know, when something's going wrong, you're getting dumped with adrenaline.

So we had to not only make it easier to fly, but we also had to encode all of those procedures. And in many cases, we've automated them insofar as we can while retaining agency with the pilot. So it was a huge undertaking. We believe that the system at the end of the day puts the pilot more firmly in control. So if you're on board that aircraft, just as you experienced, you're empowered and you're not worried about systems management. You're a mission commander now, similar to how you would feel if you were flying an F-35. So we feel like we're elevating the pilot.

Now, if you want to take the pilot out of the vehicle, same story, right? You're elevated to a mission commander flying that very same machine with this high level of intelligence and, you know, million-plus lines of code embedded in the aircraft. Now taking it one step further to your question about, you know, how do you think about the features and what are they? This has been the interesting thing over the last several years. We literally have product-market fit with everything that flies there. Every single end use case is in our backlog, both civil and defense, and every one of those customers comes to us and says, here's my specific mission. Can you build software application layers on top of SkyOS to help me do my mission better? And the answer is always yes. It's just a question of how do we sequence these features in these applications.

This is the hardest maneuver in all of aviation. It is when a helicopter loses the engine and can no longer fly level flight or climb. Contrary to what some people believe, if a helicopter loses the engine, you're not automatically dead. You can immediately enter into this configuration known as an autorotation. In the case of the Robinson R66, you have 1.8 seconds to react to noticing that the engine is no longer there. If you don't respond in 1.8 seconds, the main rotor continues to try to spin and

push air down. It can't because it's blocking the torque of the engine. Therefore, it stalls. And from that point, you are a fridge. If you do respond in 1.8 seconds or less, you are then in a very controlled, potentially somewhat free fall, right? It's not as fast as a free fall, but you are making your way towards the ground very rapidly. As you make your way towards the ground, you're building up kinetic energy inside the rotor, the rotor system. So, think maple seed, right? So, maple seed spins to the ground and it spins faster and faster. Here, we don't increase the rate of speed past a certain point, but you've got a lot of energy being built up into that rotary disc. If you are a good enough pilot that you can take that kinetic energy and right before you hit the ground, convert it into a puff of downward thrust, you can cushion your landing. Do it too soon, you use your energy, you get to zero air speed, and then you proceed to fall further. Do it too late and you've already crashed into the ground. So, needless to say, you have to do it right and you only get one shot at it. And this is on the back side of the 1.8 seconds. And to do everything I just described is so difficult and dangerous to get your license to fly a helicopter. The FAA no longer requires you to demonstrate this capability because so many people have died practicing it.

Yeah. Needless to say, this was the very last thing that we went out to achieve with Sky OS because you don't want to get this wrong and you want to refine the operating system in so far as you can. Nobody had ever attempted it, but it was necessary because if we could not solve this scenario with Skyos, then it would mean the mechanical controls would have to be still in the cockpit. And in the event that you lose the engine, let go of Sky OS and go grab a whole set of mechanical controls and carry with it, you know, all that training and complexity that I just discussed. So instead, we said, okay, Skyos has got to manage this situation. And what happens now, if you were to lose the engine in a helicopter, there is no immediate reaction required from the pilot. Skyos immediately puts you into an auto rotation faster than the human can even perceive or hear the horn light. So you're in the auto rotation and from there it is up to the pilot where they want to land. So you're literally just steering the aircraft left or right. It's never going to let you exit what's safe as you continue in a controlled fashion. And before you get to the ground, it's up to you as a pilot to dictate how much forward air speed you want to set down with. And we've had a number of people, it's probably north of 50 now in our conforming simulator. And we can teach you how to do an auto rotation your very first time. So it's like a non-event for non-pilots. So we've trivialized, you know, the most difficult emergency maneuver in all of aviation today.

That is the pinnacle.

Yeah, the challenge here is the system isn't really, there's some pretty significant gaps in the training process. You'll find a lot of times tour operators and tour pilots are pilots building time because they're not yet hirable by some of the more complex missions like air medical services or offshore oil. Those jobs pay better, but they're waiting for the pilot to quote-unquote build time, get hours in their log book, and prove that they're good pilots. Even worse, I would say, is flight instructors are building time, generally speaking. So, you have people that just finished their license, got through their commercial, and then became certified flight instructors, the lowest time people teaching the new pilots. And that's incredibly dangerous because you have someone that's, you know, a couple hundred hours of experience and flight time with someone who literally has no idea how to fly and is trying to figure it out and every once in a while is doing really the wrong thing. So there's a lot of accidents where the student does the wrong thing and the instructor, even if they're a fantastic pilot with thousands of hours, you know, maybe there's not an opportunity to recover. And again, get back to this point because why is that the case? They're trying to build time in order to graduate into higher paying roles within the industry. It is true that the pilot has a greater effect on the safety of the flight than any other factor. Certainly as it relates to a helicopter and the bulk of the accidents, it's like 70, 80% plus for both airplanes and helicopters are again related to what the industry calls pilot error. We don't see it as pilot error. We see it as really an industry error. The technology in these cockpits has not gotten simpler over time. It has gotten more complex because nobody's done a clean sheet approach to how do we redefine the human machine interface as we have to make it safer and easier. Instead, as new capabilities have come to bear over the last several decades, new boxes, new systems, and new buttons. And so now you have the pilot that would be better served as an octopus that's supposed to go between all these different things. And again, when something goes wrong, well, it's a cascading set of problems. And that's why these statistics are what they are.