Unmanned

T.H.ursdays with Tom Hill
 

This weekend I saw the movie Her. The movie concerns a soon-to-be-divorced writer who forms a friendship with his Operating System. The idea is the Operating System, an application you can carry with you like your phone, becomes this person’s friend, confidante, and relationship partner. While that was all cool and the movie was very beautiful, those features have little to do with Karlene’s website and its focus on aviation. That is, unless you think what else that Operating System could do other than be your best friend. 

 

How about flying your airplane?

A friend of mine, who is one of the first graduates of the of the Remotely Piloted Aircraft (RPA) course at the Test Pilot School, asked me, “Could UAVs [Unmanned Arial Vehicles] ever get to the point of replacing pilots?” Well, I reminded him they already are. What he was really asking was could the systems we build be robust enough to pilot passengers from one destination to the next? I replied, “Yes, it’s only a matter of time.”

We’re just at the beginning of a major revolution in automation in our societies. It wasn’t long ago when cars were made by actual people. Nowadays, most assembly is done by robotic processes. Few people really machine anything any more. They more often CNC the part they need. The part is perfect and identical every time, better than any master craftsman could ever achieve. The hardest part of the CNC process is programming the brains of the machine correctly.

That’s essentially the challenge nowadays for fully automated passenger flights: making the computer systems smart enough to exceed what the average pilot could do. Let’s use “see-and-avoid,” the basic tenant of VFR flight, as an example. Except in very special circumstances, UAVs still can’t operate in the National Aerospace Structure unless there’s guaranteed separation of aircraft. Without a see-and-avoid capability, UAV’s can’t operate in most of our airspace without special permission. So, the question becomes, “How can we emulate see-and-avoid in our unmanned aircraft?" Well, that’s a great question because there are no defined standards other than the physical characteristics of a medically qualified pilot. 

 

As I participated in several of these see-and-avoid technology development projects, I noticed the capability level the regulators were nudging the manufacturers toward mostly exceeded what I could do with my Mk 1 eye-ball. They were making the UAV see better than I could. But, making it see well wasn’t the real challenge. Making it know what to do was the challenge.

Really, the hard part is making the UAV do the right thing based on what it sees - in other words, making the brains of the UAV be smart enough to emulate what a pilot could do. Most of the projects I worked on were not very adaptive. There were look-up tables, references to canned circumstances, then built-in responses to those scenarios. 

 

When the UAV would see the intruder aircraft, it would assess its database of various scenarios, then execute the prescribed response. It was all very canned. While all this was pretty cosmic looking, on the Bloom’s taxonomy of learning scale I introduced last week, this level of “knowledge” is relatively low. It might be at the application level. More likely, it’s still at the remembering level where all you do is regurgitate the proper procedure. Using this, the UAV would encounter the intruder aircraft, assess it, look up its table of different scenarios, then execute the response directly connected to that scenario.

The innovative thing in the technical world nowadays is computing systems that appear to be adaptive. These are systems that are able to interpret scenarios and create solutions that previously didn’t exist - and do it really quickly. That’s what’s showcased in the movie. In the crucible of a social situation, the Operating System constantly adapted to new inputs. It created interesting outputs. It made for an interesting movie as it matured.

Back to my friends question, “Could UAV’s…?”

There are two ends of the computing spectrum to create a system that could replace pilots. The first is like I described in the see-and-avoid development. You could populate the brains of this UAV with all the scenarios and situations we could possibly conceive of. With that knowledge, the UAV would operate reliably and predictably for each one of those scenarios every single time. But, what about situations the UAV wasn’t programmed for? 

 

Remember, with this type of programming (the look-up table way to computing) it’s operating at a pretty low knowledge/learning level. This means it would predictably fail in unexpected and un-programmed situations. Hmmm....this sounds like what happens when human crews encounter situations for which they are not well trained!

Now consider the adaptive system that’s able to interpret scenarios and glean information beyond what’s enclosed in its database. Instead of accessing a simple look-up table, it would assess the situation, develop a response, and execute it correctly. The situations it could reliably handle would be much broader than the other system. 

 

In almost every way, this design approach is better except for at least one: creating such an adaptive system is not easy. In the same way you have to be committed to your human crew training—to a higher level than reinforcing rote memorization—you also have to train it to the circumstances of flight in order to improve its level of knowledge. Can all this happen? Yes, I believe it can, and in the not too distant future. Will it cost a lot of money? You bet. If anything is constant with these computing technologies it's that they are expensive.

Here’s the thing I find ironic: The same bean counters who might minimize pilot training in order to save a buck are likely to be enamored with the perceived savings they would get with automated systems, without realizing that automated systems carry a support requirement just like we human pilots do. While I do think advanced computing technologies are in our future, I’m not really afraid of them. 

 

 I look forward to their arrival in the same way I looked forward to the advent of the personal computing age. I believe these technologies don’t have to happen at the expense of our current needs; it’s not an either/or situation. But, it makes no sense to me to require our automated systems to work at high knowledge/learning levels and not require the same of our pilots. Instead of waiting for the advent of new technology, we can use our current resources and have mostly the same result. It’s a matter of commitment.

What Do You Think?

Cheers
Tom