[PDL]

The most prevalent autonomous vehicles currently operating in the world are airborne ones -- drones. From an engineering and control theory standpoint, an autonomous winged aircraft is no harder than a drone and drones don't need millions of dollars of avionics to fly. In fact, for many phases of flight, even passenger aircraft are autonomous vehicles in using an autopilot for both cruise and landing.

Not quite true here. Most of the armed drones are remotely piloted from Holloman AF base in NM. You need a pilot to fly the "drone" or UAV in circles around the target and another human set of eyes to lock onto the target. If I recall correctly most target strikes are verbally authorized before both members of the team fire the weapon. I also remember reading about the high incidence of PTSD on "drone" pilots because of the stresses of the job.
Inside a US Control Centre for Drones | Military.com

Also note, the "drones" do not carry humans, commercial and military aircraft carry humans and the avionics on those aircraft do cost millions to build, test and maintain. With unmanned aircraft you don't need the life support systems but then you don't have thousands of long haul drones the size of a 777 in the air at one time either. Check out this cool map.
Flight Tracker | Flightradar24 | Track Planes In Real-Time

---------- Post added 07-23-18 at 12:07 AM ----------

Having a human pilot "just in case" something goes wrong presumes that pilots don't cause crashes (and that AI can't handle emergencies). Both ML370 and Air France 447 show that human pilots are a danger. Figures from Boeing suggest that 50% of crashes are pilot error although it's hard to know the numbers of crashes averted by human saviors (and the percentage of averted crashes that an autopilot could have prevented, too).

I think using ML370 is a false flag here, since the airframe has not been located it can't really be attributed to humans causing the incident although a system failure is difficult to imagine. AF447 was due to very poor training on Air France's part and to issues with the design (and replacement schedule) on the pitot tube heaters. On AF447 when the "regular" captain returned to the flight deck he did try to save the aircraft but there was just not enough time for him to do it. Yes, there have been very few incidents where the machine was the source of the problem. Even OZ214 was due to pilots not knowing how to fly the 777 manually and the reason they had to fly manually was that the ILS was off-line.

So here is a situation where even manned aircraft with automated systems need to communicate with the ground in order to survive. What would happen if a autonomous aircraft full of people could not land because it can't talk to the ground equipment. I would hope that there was a trained and qualified human at the controls fallible or not.

---------- Post added 07-23-18 at 12:16 AM ----------

There was a military research project done a couple of decades ago on using neural nets to automatically fly a damaged aircraft. The project even installed the system in an F-15 which was then test-flown with simulation of combat damage by disabling different control surfaces. The neural net could learn how to fly are incapacitated aircraft within a number of seconds simply by manipulating the controls and monitoring how each worked (or failed to work). An AI could be readily trained on every emergency that ever occurred and remember it flawlessly if it came up.

It may be true that adding enough 9's to the reliability of an autonomous aircraft control system is a concern although the number of 9's in the reliability of human pilots may be a lot lower than people think.

So puzzle me this? Why if autonomous systems are so effective, why does the military still train humans to fly their aircraft. And yes, even the Navy has tested autonomous aircraft that can land on a carrier, but they have yet to demonstrate that a autonomous aircraft can land on a carrier at night, in the rain with 30 foot swells. Some humans have been doing that since carriers started running night operations.

---

Last edited by PDL; 07-23-2018 at 12:08 AM.

[dbs]

Hi PDL

Why do you assume that all autonomous vehicles are ONLY for military use ????

Dave

[Liney]

It isn't as though having a human at the helm of an airplane is a recipe to fix all ills. Reading the story of what happened with Air France Flight 447 is scary in the sense that the first thing they teach you in flight school is to point the nose down when you go into a stall, but somehow when humans are under stress and the plane is stalling they try to pull the nose up.

There has been a concern raised in several aviation circles for a while now that pilots are not being trained to fly the aircraft, but are being trained to diagnose the problem first. This is not a recent event, there was a Ansett jumbo landed with the nose wheel up in 1994 and the investigation found that the flight deck crew had been concentrating on working out what the horn they were hearing was for and didn't fly the plane first.

The latest concern is that with the level of automation on flight decks the problem itself may be hidden, which was one of the conclusions on AF447. The flight deck crew spent so long looking for the fault which was giving the instruments the incorrect indication that they forgot to fly. I believe it is still the case that every aircraft must have a set of pitot-static instruments regardless of how computerised the flight deck is, and by concentrating on these the AF447 fault may have been more quickly identified.

As for drones, in my opinion the most significant difference between an autonomous aircraft and an autonomous car is the amount of traffic it has to contend with. OK the car only has two dimensions to care about not three, but the separation distances between the car and the hard things it has to not run into is considerably less than airborne traffic for a drone.

[Racer X 69]

It isn't as though having a human at the helm of an airplane is a recipe to fix all ills. Reading the story of what happened with Air France Flight 447 is scary in the sense that the first thing they teach you in flight school is to point the nose down when you go into a stall, but somehow when humans are under stress and the plane is stalling they try to pull the nose up.

I think too of the whole Toyota stuck accelerator fiasco. Clearly there was some defect in these Toyota vehicles, but if you just hold the brake pedal down, the car would still stop, even if the accelerator is also pegged to the floor. Malcolm Gladwell did a whole podcast on this a couple of years ago and came to the conclusion that the drivers tensed up, thought they were pushing on the brake and in reality were actually pushing on the accelerator and the faster they went the harder they pushed the gas pedal down, thinking that it was the brake.

Humans really have an overly inflated sense of how good they are under stress, but few of us perform all that well and in fact machines (lacking emotions and fear) are probably better at dealing with those situations, assuming adequate programming ahead of time.

I always wonder why those who find themselves in a car with an accelerator “stuck” wide open don’t simply take it out of gear?

The ECM will prevent the engine from over revving (although it does sound bad when it is up against the rev limiter). Then one simply applies the brakes, brings the car to a stop and turns it off.

[Rondec]

I always wonder why those who find themselves in a car with an accelerator “stuck” wide open don’t simply take it out of gear?

The ECM will prevent the engine from over revving (although it does sound bad when it is up against the rev limiter). Then one simply applies the brakes, brings the car to a stop and turns it off.

Of course. Put the car in neutral. Take your foot off the pedal and try again. Pull/push the emergency brake. There are all sorts of things you can and should do in such a situation, but when people are under stress they just don't do those things. They have their foot on the pedal they think is the brake (but is actually the accelerator) and when the car doesn't slow down, they just push it harder till they crash.

I don't know that autonomous vehicles are a solution (although they may be partly), but what I do know is that most of us actually tense up and perform very poorly under pressure.

To use a sports analogy, quarterbacks often will have very high completion percentages when they have good protection, but when they are "under pressure," this often drops significantly. This performance hit is true for all of us, if we are honest.

As I think about it, one of the negative effects of autonomous vehicles would be that we actually have less experience in driving, both normal and difficult conditions, but would only be called on to pilot our vehicles in difficult situations where systems were failing. That could be problematic.

[PDL]

Hi PDL

Why do you assume that all autonomous vehicles are ONLY for military use ????

Dave

Where did I say that autonomous vehicles are only for military use?

I was referring to autonomous aircraft, which is where the thread has been for the last few posts, no plain old generic vehicles. Aircraft are not cars and it is much more difficult to take off an land than to fly. Heck most modern aircraft are somewhat autonomous now during the flight phase due to autopilot functionality. It is the takeoffs and landing that can get you killed the quickest and for the most part that part of the flight is under human control.

Name one human carrying autonomous aircraft and I mean an aircraft carrying a human passenger or passengers without a pilot. Even the so called autonomous cars out today run on restricted routes and under current law they require a steering wheel, accelerator pedal and break pedal. If you watch the video link that I posted before, you will note that all of the cars listed have humans who can take over.

The Waymo cars are apparently leading the pack with their autonomous cars, but they are being introduced in a limited way in Phoenix AZ.
<snipit>
Waymo self-driving cars will be used in a ridesharing service launching in Phoenix, Arizona, later this year. Vehicles will initially be limited to designated areas of the city, but Waymo plans to expand the area of operations and possibly add more cities in the future. Arizona has among the most permissive autonomous car laws of any state, which may have played into Waymo’s decision to launch its service there, rather than in its home state of California.
</snipit>
https://www.digitaltrends.com/cars/waymo-self-driving-cars-reach-8-million-m...-public-roads/

Please note that not only is AZ used because of the "permissive autonomous car laws" but it does not have the really bad weather found in some parts of the country. AZ => Sunbelt.

Let me again state that I would really like to have autonomous cars that I could "rent" or "request" at my convenience. But having grown up in a state that has very bad weather during most of the year, I think that those people who are not living in the Sunbelt should demand more rigorous testing and development of the requirements for use by the general population.

[photoptimist]

It is the takeoffs and landing that can get you killed the quickest and for the most part that part of the flight is under human control.

That's because human beings really suck at landing -- the body and brain don't have the sensors, the control theory math, or the multitasking needed to optimize a smooth landing. Once you experience an autoland in a 777, you'll never want to go back to human pilots.

[clackers]

I think using ML370 is a false flag here, since the airframe has not been located it can't really be attributed to humans causing the incident although a system failure is difficult to imagine.

"False flag"?

(Laughs)

So humans didn't switch off the beacons, and make two smooth course changes afterwards!

Fascinating, PDL.

[dbs]

Hi all

I honestly feel that human and autonomous vehicles ( mainly cars ) cannot share the road at the same time.Second quessing all the time will get you killed.
PDL maybe you should re-read some of your earlier posts to know how I came to that conclusion'.

Dave

[onlineflyer]

That's because human beings really suck at landing -- the body and brain don't have the sensors, the control theory math, or the multitasking needed to optimize a smooth landing. Once you experience an autoland in a 777, you'll never want to go back to human pilots.

Tell that to the pilots who have landed disabled aircraft.

[PDL]

That's because human beings really suck at landing -- the body and brain don't have the sensors, the control theory math, or the multitasking needed to optimize a smooth landing. Once you experience an autoland in a 777, you'll never want to go back to human pilots.

Wait, wut?
Humans are at the controls for the majority of takeoffs and landings, I think you have the idea turned around in this case.
From a discussion on FlightAware

"Assuming that takeoff= landings minus crashes, and crashes are very rare, essentially zero on any given day, we can assume there are 106,000 takeoffs worldwide in 24 hours.
That works out to about 4400 takeoffs per hour averaged over a day. Since it’s worldwide and based on 24 hours, the day/night factor is sort of balanced out.
I believe the flightaware data includes commercial passenger flights as well as general aviation cargo, private aircraft registering flight plans including corporate jets, but not military flights."

I would imagine that the total percentage of "Autoland" landings is quite small. Just how many accidents occur per day? From the 50's to 2010 it was .04598 crashes per day. How many planes crash every day? - Quora

Not bad for something that humans suck at eh? I suggest that you look at this site.
http://www.pbs.org/wgbh/nova/space/how-risky-is-flying.html
Take note of the first chart - Risk of death by Year
automobiles 1 in 700
aircraft 1 in 1,206,000
and to think - humans suck at landing...… Really

---------- Post added 07-23-18 at 04:00 PM ----------

"False flag"?

(Laughs)

So humans didn't switch off the beacons, and make two smooth course changes afterwards!

Fascinating, PDL.

All the existing evidence appears to show that in this case the humans involved were responsible for the aircraft loss, the use of ML370 as an example in this case is misleading since we really do not know what happened. We can speculate all you want, take educated guesses, blame aliens, hijackers and call out a myriad of other scenarios in the case of ML370. However, from my perspective, if ML370 was downed because of the overt actions of the pilot, that is not a failure, that is/was the intent. Failure and intent are two different things. Remember the aircraft was on autopilot and from all the available information the autopilot, transponder and other systems used to communicate with air control was intentionally turned off. In this case I would not classify the hull/human loss as a failure but as a success. Harsh, yes, but it appears as if the person(s) completed what they wanted to do.

[MarkJerling]

I suspect we'll need humans for landings like this for a while longer!

You Tube

You Tube

[PDL]

Hi all

I honestly feel that human and autonomous vehicles ( mainly cars ) cannot share the road at the same time.Second quessing all the time will get you killed.
PDL maybe you should re-read some of your earlier posts to know how I came to that conclusion'.

Dave

While I may agree to a certain extent what will be one of the most difficult aspects of autonomous vehicles on the road (not talking aircraft or watercraft here) will be the coordination of human controlled / autonomous vehicle cooperation. Some where along the line it will be regulated to a some degree. Who will draft/enforce the regulations is the question. Government, Facebook, Google, Uber, the car companies or some new startup that only drives on sunny days?

---------- Post added 07-23-18 at 04:16 PM ----------

I suspect we'll need humans for landings like this for a while longer!

Absolutely, plus add in aircraft carrier operations in foul weather.
For more exciting landings just go out to YouTube® and lookup Wellington NZ landings. Thankfully when I landed/took off from there it was a really nice calm summer day.

[pschlute]

Both ML370 and Air France 447 show that human pilots are a danger.

As a percentage of all flights, incidents like these are infinitesimal

although it's hard to know the numbers of crashes averted by human saviors (and the percentage of averted crashes that an autopilot could have prevented, too).

And how would an autopilot deal with a stuck landing gear or a flap or slat that wont retract?

Autopilot may well be used for 95% of the time the aircraft is in flight but it is a tool of the pilot. It is not capable of managing the aircraft at all. It will only follow the instructions given to it.

[MarkJerling]

Absolutely, plus add in aircraft carrier operations in foul weather.
For more exciting landings just go out to YouTube® and lookup Wellington NZ landings. Thankfully when I landed/took off from there it was a really nice calm summer day.

I've landed there in foul weather several times. One landing, specifically, was a white knuckle ride. We approached from the South. Windy day, thick and low cloud. When we eventually broke through cloud cover, we were lower than the end of the runway, just above the waves. Pilot had to open up the throttle and climb the plane to get to the end of the runway. All the while, the whole thing was pitching up and down and left and right. When they eventually dropped onto the runway it was at an angle. Hard drop and bounce. We'd just come back from Melbourne where the kids had been on a rollercoaster. They were thrilled. Said the plane ride was "fun" like the roller-coaster at Luna Park!