Paul DeHerrera spent a career making avionics instruments to deliver an ever-swelling river of data to pilots to help them safely navigate the airways.
Now, after retiring as CEO of longtime Tucson avionics maker Universal Avionics following the company’s sale, DeHerrera is working on a second career with a new mission: Helping pilots manage and make sense of all that data, using rapidly emerging artificial intelligence technology.
Incorporating artificial intelligence technology into existing industries and equipment in the name of efficiency, efficacy and safety is an emerging trend — one that is not lost on Tucson startups like DeHerrera’s and the University of Arizona, which is working to beef up its AI offerings.
AI tech, broadly, are systems that mimic human intelligence by interpreting and learning from data to achieve specific tasks or goals.
DeHerrera, 66, believes there’s a growing need to incorporate AI into aircraft cockpits.
“When you’re flying as a pilot, you’re inundated with an incredible amount of information, and most of it is very valuable,” said DeHerrera, a longtime pilot as well as an electrical engineer with an MBA.
He cited instruments that monitor the health and status of aircraft systems, flight-management systems, terrain-awareness systems, communications systems and radars and air-traffic monitoring systems.
“There’s an incredible amount of data coming into the cockpit ... Even in my light aircraft I’ve got exhaust-gas temperature information coming in for each cylinder,” he said.
But all that data isn’t much good if it overwhelms the pilot.
“As a pilot, you have a limited amount of time, you can’t just sit back in your easy chair and process this data, and there may be something very important that you could be overlooking because of the volume of data — while you’re focusing on the engine, there may be something about weather that comes up, or about traffic, and those things can happen very quickly.”
And so, Innovative Avionics was born, launched by DeHerrera and Universal Avionics’ former owner, Ted Naimer. The pair joined forces after Israel-based Elbit Systems acquired Universal in April 2018.
The startup company is initially working to develop pilot assistance systems using artificial intelligence and “machine learning” — an application of artificial intelligence allowing devices to essentially learn from data inputs without human programming — to rapidly analyze all the navigation and other data and prioritize it for pilots.
“Artificial intelligence and machine learning is a really good way to do this because the computer can be looking at all these things that are happening and can make a decision on what’s important to the pilot and what it should bring the pilot’s attention,” DeHerrera said.
He said the company is looking to initially use AI to develop software that can help pilots go through checklists commonly used for different stages of flight, such as startup, takeoff, cruise, descent and landing.
DeHerrera envisions an initial system using a digital assistant — much like Siri or Alexa — that can use voice recognition to go through checklist items with a pilot, much like an airliner co-pilot.
The system could be used with an app on a smart phone or tablet computer, which are already commonly used by both private and professional pilots, he noted.
THE SKY’S THE LIMIT
Beyond informational and advisory uses, AI could someday extend to flight controls and eventually, autonomous flight, DeHerrera said.
Autopilot controls were invented not long after airplanes in the early 1900s and their use is now commonplace during cruise phases of flight.
While many large aerospace companies are working on incorporating AI into avionics, fully autonomous flight is likely years away as regulators try to sort out the technology and safety issues.
DeHerrera said sophisticated autonomous control systems already have emerged for ground vehicles, citing TuSimple, which has been testing self-driving truck technology at its Tucson test center since 2017.
“It’s actually more difficult to drive that truck on the road (autonomously) than it is for us to fly an airplane, because airplanes are in a very controlled environment,” he said.
However, two fatal crashes of Boeing 737 Max jetliners have been blamed partly on an automated system designed to counter a pilot’s propensity to ascend too steeply.
The crashes have been blamed on the system along with poor training, and the pilots involved reportedly struggled to find solutions in manuals as they were going down.
DeHerrera said AI systems could be designed to quickly offer pilots emergency solutions verbally — perhaps helping to prevent crashes like the 737 Max mishaps.
“We can access data much faster, and with AI, we can sort through that, based on the phase of flight,” he said. “So when the pilot asks for an emergency checklist for so-and-so, bam, we pull up the checklist right away.”
For now, DeHerrera said, Innovative Avionics is in the process of building an initial prototype, which they hope to start testing in the fall, with the help of a Ph.D. expert in artificial intelligence and a software engineer.
The company plans to start in the experimental aircraft industry — which includes research and training aircraft, air-racing planes and home-built kit planes — to get valuable pilot feedback on its system interface, DeHerrera said.
A University of Arizona official involved in forging links with the business community said he expects more local companies to spring up based on AI capabilities.
But given that AI is being developed as a tool to advance technology in a variety of industries and existing companies, its impact may not be readily visible, said Stephen Fleming, UA vice president for strategic business initiatives.
“This is a real hardcore AI problem, and what’s interesting is, we’re seeing that everywhere,” Fleming said of Innovative Avionics’ efforts.
Besides TuSimple, Fleming cited the self-driving mining trucks built by Caterpillar and tested near Tucson.
“We’re seeing it in aviation with Paul’s company, we’re seeing it in medical imaging, every time you turn around it’s leaking into other areas, it’s becoming an underpinning technology,” Fleming said.
He noted that the UA has been boosting its AI offerings in the Department of Computer Sciences, but those graduates may end up working in any industry.
Meanwhile, the UA Department of Aerospace and Mechanical Engineering is currently looking to hire an associate- or assistant-level professor with expertise in AI.
“Before, you didn’t work with AI unless you were a researcher,” Fleming said. “Now if you’re going to build bulldozers you have to know something about AI.”
