David Simpson always dreamed of being a pilot. When he was a child in the U.K., his father was a member of the Royal Air Force, and he used to regale the boy with stories, stoking those desires.
That dream never came true.
The next best thing did, though: Simpson became an avionics engineer. After getting his start working on Vanguard-class nuclear-powered submarines for the Royal Navy as a member of Vickers Shipbuilding and Engineering (now part of BAE Systems), Simpson took to the air to leave his mark on the Royal Air Force’s Panavia Tornado avionics just before the first Gulf War.
In 2003, he left the U.K. for the U.S., and the military for commercial avionics. Last year, after a decade at the helm of a small safety-critical-focused avionics company named Richland Technologies, the company, which he co-owned, was acquired by Mercury Systems.
Simpson now serves as the senior director of safety-critical products and systems for Mercury Mission Systems. He spoke to Avionics about his role at Mercury, the growth of autonomous systems and the difficulty they pose for regulators.
You’ve been in avionics for a while. What stands out as the most fun or interesting project you’ve worked on?
I have two, actually. We did a retrofit display for business aircraft, so basically replacing cathode ray tube (CRT) displays with liquid-crystal displays (LCDs) to replicate the exact same functionality of the old CRT display, even being able to mix and match CRTs and LCDs in the same cockpit. Then, just by adding some additional wiring into the aircraft, bringing in all these new concepts: XM weather, Jeppesen charts and maps, synthetic vision — but you still have the legacy part of the aircraft. I took that project from conception and even went on flight tests with it.
The next one is the enhanced vision systems that I've worked on: the infrared light, cameras and image processing to allow landings in poor visibility.
What about that still sticks out to you?
Leading-edge technology. Solving a problem that would have never been solved before, allowing aircraft to go into airports where they normally wouldn't be able to go and getting that kind of system through certification and seeing it help the pilots get where they needed to go was very rewarding. I worked on that from software definition through flight testing.
Walk me through your day-to-day at Mercury Systems.
My basic occupation is to ensure that Mercury is fully in line with developing safety-critical products for the commercial off-the-shelf market. Typically, Mercury System would not work in commercial avionics, wouldn’t work in real safety-critical system, so I’m assessing new products, new technologies to keep everything in line with the safety-critical standards. I'm working within Mercury to ensure that we get things done across the whole company and standing up the safety-critical expertise within Mercury and the certification expertise and allowing the military products to be deployed for commercial.
You know the FAA and EASA certifications. There's a huge difference with getting something — I put this in air-quotes — certified on the military platform than there is going for a full FAA or European EASA certification.
Is there anything you all are working on right now that you can tell me about?
We’re working on small form-factor certifiable computers targeting the unmanned aerial vehicle (UAV) market. I've got a program that I can't actually mention, but we're on contract for a program to produce a UAV flight computer that's going to control all aspects of the flight that can be modified and is cost-effective for UAVs.
Speaking of unmanned vehicles, what are your thoughts on how they are growing and changing the industry?
UAVs basically first came on the rise in the military-only kind, but that's definitely changed. UAVs are now becoming very widespread in the civilian space and in civilian applications like marketing, general photography and aerial surveillance of traffic.
The biggest expanding market and innovation is in the commercial space and now we're seeing the advent of the unmanned taxi style. That's where people want to take it, so you can just hop into one these things like an airborne Uber and it'll take you somewhere, and there are many companies that are actually even prototyping designs.
The interesting thing for me is: how is certification working on this stuff? Now we've got no pilot, and a lot of things previously mitigated by having a pilot in the loop are no longer mitigated like that. So, from the certification software technology point of view, we are exploring at the moment, and Mercury is really intent on grasping the opportunities there and becoming successful in that market. The autonomy of the aircraft is a paramount strategy for Mercury now.
With all the additional automated functions on modern aircraft that are being developed, what does that mean for the back-end in terms of required processing power?
Yeah, the processing required is obviously a challenge, but we know that's going to be overcome. We're moving into the real high-end and field-programmable gate array process, general-purpose graphics processing unit process to get the power that's needed for that — and, obviously, the power is just increasing and increasing. So, you know that we'll get there.
So again, how is the certification going to be affected by this? What you see there is like an enormous increasing complexity of the system, complexity of requirements, complexity in verifying these requirements and validating the requirements. That's where we're looking now with partners to see how it is going to get handled.
We know there's movement within certification bodies to recognize that this is government, and I think we've already started working groups to look at how authorities and the traditional ways of doing certification are going to get modified to come along and help the industry get what it wants from the technology to be able to safely bring it into the public aerospace.
Do you have any thoughts on what that adaptation should or will look like?
The traditional mechanisms are not there at the moment. The traditional ways of doing it — a DO-178 for software, the DO-254 for hardware — is not enough to try and do it that way. It explodes the amount of work it needs to get done and everybody who's ever been involved with a certification knows how expensive it is to do it on today's traditional systems on a traditional fly-by-wire system; the certification costs are probably five times the development costs.
Adding complexity into that system — because it's still a very manual system, although there are many tools that can be used within the processors — there's still a lot of manual processing to be done. It involves a lot of people doing a lot of stuff like documentation and actual physical reviews of that documentation and audits of that documentation. So, when we think about that and adding the kind of complexity that AI brings then it really shows that a leap in the certification technologies is required as well.
My feeling is that the rule-based learning part of AI could be utilized in the certification aspects. That's what we've got to start looking at as engineers. It’s great to develop this, but are we going to make it safe, and are we going to develop the technology to support the certifications along with this technology that's given us the systems we want?
So, using the system you’re certifying to certify itself?
Everybody’s fear with AI is robots building robots. So, you've got to control that, obviously. You've got to have some aspect of control. Certification is a rule-based business. To certify something, you've got to have a rule base, and that does fit into certain aspects of the AI. If you can build a rule-based certification system, you've got to somehow replace the endless hundreds of thousands of man-hours of grinding through documents and test cases.
What do you do outside of Mercury Systems? What are your hobbies?
Well, most of my time at the moment is taken up because I’m having a house built. Other than that, I still very strongly follow the soccer scene in the U.K. I was always brought up going to soccer games there. Burnley has a team in the Premier League that's seventh; that's really good for a small town of about 65,000 people. I live by Lake Lanier here in Georgia, and I like to go out boating and fishing.
Does your job and background as an engineer bleed into your day-to-day life?
I'm trying to build my own home-automation systems and things, sticking gadgets in my boat. That's one thing I'm hoping to get in the house we're having built, to have as much of the technology as I can and smart home-type things. I started off using the Amazon Alexa systems, but I like adding on with different components — even things I might have built myself.
This interview has been edited and condensed for readability.