Autonomous Flight

Humans fail. Computers do so too. But there is a difference. We are able to improve computers a lot. While we won't be improving humans that much.

Aviation is the area where computers have been introducing the biggest changes. I would even say computers allow the modern aviation to exist. From design of aircrafts through handling the air traffic through handling the operations of airlines. Nothing would be possible today, have we turned the machines off.

It is also impossible to fly a modern aircraft without a computer. The fact is pilots today operate computers that operate an aircraft. And in most cases the planes fly themselves. Humans are needed to go through a printed checklist and handshake some verbal commands with ground staff. Still verbal, because there is no unified digital plane-to-earth interface in place.

Humans seem today like artificially introduced parts of the system. Almost unnecessary. And they fail. Human error has been by far the most common cause of all aviation incidents and accidents. Computer failures were hitting the headlines, when they occurred. But they no longer do. Today it is either a fatal mechanical failure (in minority of cases) or human failure (in vast majority of cases, like the recent YAK-42 crash). And we also have failures of communications between computers and humans, the AF447 being the crown example.

There is nothing today preventing us from introducing pilot-less planes. We have mature technology able to take off, fly and land any plane. It is just a matter of putting all the pieces together. Backup systems. Remote supervision. Cryptography securing the earth-to-plane communications. Telemetry. Proliferation of advanced ILS systems on most airports. Ultimately we will have to convince passengers that pilot-less planes are safer, because nobody will mess with the computers. In the meantime there is the cargo segment that will be on the leading edge. UPS and FedEx already have the most advanced air traffic handling systems in their Louisville (SDF) and Memphis (MEM) hubs. Technologies like ADS-B (Automatic Dependent Surveillance-Broadcast) and CDAs (Continuous Descent Approaches) have been researched and developed by those parcel shipping companies (not military nor government!). They result in increased safety, reduced fuel burn and reduced flight time, but completely rely on computers. Even the best human pilot, not assisted by these systems, would not be capable of safely approaching SDF or MEM in the crowd of hundreds of computerized planes around.

Autonomously flying planes will revolutionize private / personal transportation too. Look. It is MUCH easier to build a plane that flies itself, than a car that drives itself.  A plane does not need to follow a road, stop on traffic lights, watch for pedestrians. It just needs an air traffic avoidance system (which is much simpler than an road traffic avoidance system) and instrumented landing pads. High skills needed to become a pilot have been the main roadblock in widespread expansion of personal aviation. So has the lack of experience of many private pilots resulting in tragic accidents. But when computers take over the cockpits, anybody will be able to fly.

Looking at the timeline ahead of us, I think we will have autonomous cargo flying over our heads in this decade. And then may be it will take until 2025 for passengers to board an autonomous plane. But it will happen. And I am already lining up to be one of the first passengers on such a flight. In the end, does it really differ from riding an automated elevator to the 35th floor in a hotel or an apartment building?

Comments

  1. nstead of pilot errors, software bugs?

    ReplyDelete
  2. Yes, true, but I am sure ultimately we can make computers far more reliable than humans. And speaking of automated transport, it is orders of magnitude simpler task to build a machine that flies itself from Paris to Moscow, compared to one that drives itself the same route...

    ReplyDelete
  3. I disagree that computers will be more "reliable" than humans. There is a tremendeous amount of judgement and intuition used when flying today. Computers are able to fly the aircraft, they already do so today in most aspects of commercial flight. Pilots couple the auto pilot to their planed nav route and the computer flys the aircraft at the optimum speed for the altitude the aircraft has been cleared for. But there are numerous situations where a pilot needs to make decisions, and a lot of those decisions are based on his or her's experience and judgement that cannot be programed into a computer. For instance, on final approach to an airport a pilot might decided to fly with an additional 10 kts of airspeed due to turbulent wearther, increasing his stall margin and bleeding the excess off in the flare to landing. Small example, but those types of decisions are going on constantly.

    ReplyDelete
  4. Buck, thanks for your input! I am not saying the road to self - flying machines is quick and easy. I am saying it is easier compared to the road to self - driving machines. Actually computerized flying seems to be algorithmically easier task to accomplish. This is mainly because basically there are two stages we have to extensively care of - take off and approach / landing. And for these, we can put many supporting systems in place on the ground). Automated driving would be easy too, only if our roads had ILS-like functions built in. Which is a monumental task, far more expensive than building a couple of airports with advanced instruments assisting automated aircrafts in the two aforementioned phases.

    ReplyDelete
  5. All mentioned above non-typical situations are possible to encode. Software of autonomous flying planes will certainly be created by very experienced pilots. IMHO a big psychological barrier against flying such a planes has quite rational basis. Yes, you are right, flying seems to be much easier than driving. But stopping is one of the easiest thing that car can do. Very simple "watchdog" will be able to activate the brakes and release the door lock in a situation of software or hardware malfunction. Self-driving vehicles might be even equipped with a "red button" activating the stop procedure.
    "Stopping" an aircraft i.e. landing is the most complicated procedure of flying skill. How can a plane do that itself in case of main computer failure?
    The second risk is the terrorism. The computer-driven plane may be easier to attack and will not require suicide. Entering a self-driving plane remember that your life depends on an anonymous admin configuring the plane firewall and thousands of terrorist hackers.

    ReplyDelete
  6. Good points, Marek. My additional 2 cents: stopping is not that impossible - Cirrus has emergency parachutes in all planes they manufacture. Still there are unexplainable accidents (like the one two weeks ago in Katowice), where pilots never pulled the emergency handle. Computer would certainly do so... And as for terrorism, we are prone in so many areas... If they cannot intercept a flying Cruise missile, they will not be able to intercept an ordinary plane either..

    ReplyDelete
  7. Are you sure that we can test a software as the evolution our brains by millions of years?

    ReplyDelete
  8. Testing is a huge issue. True. But at least software, once tested, works. Which is not the case with a brain. Computers do not experience stress, emotions, do not feel sick, tired, bad, angry... They can be built so they either work or do not, but never "half" or "most of the times". Humans are unpredictable. This is the issue and motivation number 1, to move processing and decision making from humans to computers. Number two is we know how to clone one computer to another. So when the first one works, we can easily make the next million working exactly the same way. With humans, you have to repeat the long learning process and the final result is never guaranteed.

    ReplyDelete

Post a Comment