CSX expects PTC to pave way for 1-person crews; autonomous operations

OGR Webmaster posted:
Ed Mullan posted:

I'm sure that there are a few folks who were very happy that Captain "Sully" could handle his aircraft in a very unusual situation....

Ed, I'm glad you mentioned Sully. I had not thought of that wonderful example. If this had been a pilot-less "autonomous" aircraft, everyone aboard it would be dead now, along with a lot of people on the ground.

These emergency situations are the kinds of things that technology simply cannot deal with, no matter how advanced it might be. Inevitably there will be situations presented to a vehicle (truck car, train, aircraft, whatever) that the software was not written to deal with.

The biggest, most perplexing problem presented to any software engineer is that of "error checking." What happens when things don't go as they should? Error checking within a software program has to anticipate EVERY POSSIBLE ERROR that could occur and have a routine written to deal with it.

Anticipating EVERY POSSIBLE ERROR in ANY software program is absolutely impossible. This is the "dirty little secret" that the technology proponents don't want to talk about.

I test software for a living, been doing it a long time, and that is true (there is famous theorem called the completeness theorem that says that there are things a program cannot do).   Bugs are a fact of life in software, and nothing is foolproof....but that also includes human beings, too. From everything I have been told about Captain Sully (or the pilot whose name I forgot who in the late 80's made an incredible landing I believe in Iowa, with a crippled plane, steering it with the engines alone), what they did was both incredible, and also something that may not have been handled by most other pilots..so let's say instead of Sully or the other pilot, you had one of the pilots who today fly the feeder airlines,whose training is nothing like Sully or the other ex military pilots..chances are if they were flying it, they would crash. I will add that Sully, and likely in the other example, they were relying on control systems that are heavily computerized/fly by wire) and they worked, too. For all the heroics of captain sully, how many things do we have like the Indonesia Airliner, or the pilots who commit suicide by driving the plane full of people into the ground, or who fly drunk, or who screw up like the KAL flight into SF? Not knocking pilots, but what I am saying is that human pilots are not perfect, far from it,  nor is every pilot going to pull off that once in a thousand  miracle like Sully did or the other guy I mentioned. Yeah, I have heard people sneer about their pc locking up or people on here saying "They can't even make a legacy engine that works right all the time", but that misses the point, that is comparing apples and oranges, they are built for very different things to different standards

Rich's post would be correct if we were talking straightforward programming, the kind of stuff you see in routine applications, like legacy command control,tv control, your microwave oven, smart phones, that kind of software would be very difficult to create something that could handle extraordinary circumstances, when Sully did what he did he wasn't relying on a rule book (I doubt very much there is one) or a checklist, he relied on many years of experience, he relied too on intuition and inference as well. Ordinary flying can be done by such a rule based system pretty easily, they have automated landing and takeoff systems, for example, that outside of extraordinary circumstances (something like Sully faced is probably something that happens a tiny fraction of the time, prob 99.98%  of flights are going to be routine or have issues anyone could handle) can today land and takeoff a plane, and likely would make for more efficient operations, but there were all kinds of issues around using them, some were still technical hurdles having to do with the ATC system being ancient at the time, others were resistance from the industry and pilot association. 

The thing is, that kind of programming, the kind of routine, even rule based system, programming is not what would likely be up there. One of the things that has made tremendous improvements is machine based learning, where literally systems learn as they go, they do what human beings do (in different ways), as they gain "more experience" the system becomes better and better. This  isn't science fiction, there are systems out there already using this kind of technology, systems that instead of programmers going in and rewriting routines, the systems can create their own rules based on what they come across as they go. This code isn't foolproof, and any such system would need a lot of testing and burn in time before they ever were allowed to 'replace' pilots. The thing about that kind of system is they can also 'learn' from the experience of other systems, over time as more and more data comes through from all the flights happening with these systems, the systems on board an aircraft 'learn', the way pilots today learn from talking to other pilots. So instead of a programmer guessing at situations, like "hmm, plane loses all 4 engines at 1000 feet over NYC, okay, turn the plane X degrees, set flaps to why", you could have a system that "knows" from the data ('experiences' ) of other systems, what to do/try, and to adjust if it isn't working, that is the big difference.

For the record, I don't think such systems are there yet, but in the not too near future they will be. My guess based on what I know about machine learning is initially, planes will have data recorders pulling in all kinds of data, that is used in training flight systems, helping them "learn", once they get to a certain point, have passed all kinds of simulator testing, etc, they will be onboard the aircraft, operational but not in control, and burn in further, even learning from what human pilots do...and eventually will go into trial use, perhaps with cargo planes with pilots on board acting as backup. I don't think it is going to replace pilots anytime soon, but for example, the military is actively developing pilotless planes, either controlled remotely like they do drones, or even flying themselves and operating in combat,for obvious reasons (losing a plane is a lot better than losing a pilot, if remotely controlled), plus a remotely controlled aircraft doesn't have to worry about a pilot in a dogfight getting hurt by high G forces. 

Like I wrote in another post, this has real world implications no one is really addressing, I think those who promote technology and automation and claim it is no big deal, that plenty of 'new jobs' will be created, aren't addressing the point, nor are those who scoff at what technology can do and claiming you can't replace a human being, the one thing history has shown is that both aren't true. It is true that the first phase of technology helped created more jobs, and the next ones did, but what we are seeing is different, a lot of jobs are disappearing and being replaced by very few others..and to me no one is facing this reality from any end of things.....

 

The person who dies with the best toys dies a happy person

That poor ole engineer who snaps a knuckle on Sand Patch is gonna have one heck of a job, isn't he. Oh well, the car shops in Cumberland can send someone out. Might only take a few hours or so. Providing there is not  a foot or two of snow on the ground.

Getting back to Rich's point, hard to write a program for the unknown, ain't it.

Ed

Landsteiner posted:

"This is the "dirty little secret" that the technology proponents don't want to talk about."

There aren't any secrets about this.  Occasionally even the best technology will fail, as you say. But airlines are a bad example in my view.  Highly trained, highly paid pilots plus extraordinary technology have made major airline disasters vanishingly rare.  We would be idiots to mess with that, except by adding additional technology and still having at least two pilots on any commercial flight. It's working.  End of story.

However, the current technology wedding cars and drivers is failing utterly and completely.  30,000 deaths per year, 100s of billions of dollars in medical expenses and property damage, not to mention incalculable suffering,  demonstrate that the current model is an abysmal failure.  Unless we can employ highly trained and disciplined drivers and high technology, as we have now on the airlines, nothing will change.  We are not going to make 100 million+ drivers competent to drive, attentive to the road, never drinking alcohol or doing drugs.  The only solution to these millions of preventable deaths over a lifetime is better technology, not better drivers.  It's one thing to have a few thousand great pilots paid $100-300,000 dollars per year.  We're not going to be able to do that for 200 million automobiles, are we? 

That raises a good point, that comparing airline pilots to people driving cars isn't a direct comparison. Besides the fact that pilots are well trained, aircraft today are heavily automated as I pointed out in a new post, even if the pilot is making the decisions, they are doing so aided by, for example, computers controlling the flight surfaces based on their input, the idea of 'stick and rudder controlled directly' isn't true with modern aircraft, things like the modern fighters, the stealth planes, and even commercial aircraft cannot be flown totally manually per se (or at least some of them), computerized systems can make rapid adjustments that human beings can't, so they 'enhance' the ability of the pilot.  I think the current model works, though you can argue that when problems happen often it is pilot error or human error of some sort, as rare as they happpen.

While thanks to the design of modern cars, between safety equipment, better structural design, things like ABS, lane detection and object detection/emergency braking applications, the actual death rates are down (30,000 dead, yes, but that is with driving miles many times more than they were 40 years ago; the death rate/100k miles is sharply reduced from the early 70's, when with less driving, 50,000 people died). One of the largest single causes of accidents and deaths is still DUI/DWI, despite all the increased penalties and such when groups like MADD forced change from lax laws, even though the overall rates have dropped.

Another reason for automation is outside really rural areas, most suburban and even exurban places, where most people live (and that trend is continuing on, by the end of this century some estimates are that only 5% of the people will live in rural areas) are saturated with traffic, and the current thing of building more roads and expanding existing ones doesn't keep up, peoples commuting times, especially where they drive to work, have increased tremendously since the 1970's, and one of the reasons is that highways cannot handle anywhere near their theoretical capability, because of the way people drive. Ever drive down the highway where someone in in the left lane at 30mph, or worse, a group of cars side by side are driving the same (slow) speed, or the person hunched over the steering wheel, afraid and constantly braking...and what about the large increase of us getting older, autonomous cars can mean older drivers keeping their mobility while also to be honest, without the issues that come along as people age, loss of reflexes, eyesight, etc?  Autonomous cars can help more than a few issues on the road, if they get the technology right. Sure, people point to where autonomous cars have caused accidents or even fatalities, but these are still prototypes (uber and the rest are jumping the gun, to say the least), but guess what, how many people died in the early days of airplane flight, both in the wright brothers era, and even early in commercial aviation? How many people died perfecting the automobile, or for that matter, the steam engine on railroads?  I am not saying driverless cars are ready to take over, or should, same with planes (or trains..hey, planes, trains and automobiles, should make a movie called that), just saying that new technology always has problems, has risks, and as time goes on the risks fall off as the rewards come into play, and eventually is adopted (or fails, too). 

The person who dies with the best toys dies a happy person

OGR Webmaster posted:
Ed Mullan posted:

I'm sure that there are a few folks who were very happy that Captain "Sully" could handle his aircraft in a very unusual situation....

Ed, I'm glad you mentioned Sully. I had not thought of that wonderful example. If this had been a pilot-less "autonomous" aircraft, everyone aboard it would be dead now, along with a lot of people on the ground.

These emergency situations are the kinds of things that technology simply cannot deal with, no matter how advanced it might be. Inevitably there will be situations presented to a vehicle (truck car, train, aircraft, whatever) that the software was not written to deal with.

The biggest, most perplexing problem presented to any software engineer is that of "error checking." What happens when things don't go as they should? Error checking within a software program has to anticipate EVERY POSSIBLE ERROR that could occur and have a routine written to deal with it.

Anticipating EVERY POSSIBLE ERROR in ANY software program is absolutely impossible. This is the "dirty little secret" that the technology proponents don't want to talk about.

Rich I can not help but think this is a bad move.Trains are to big to do this with.This is a very bad gamble.There to many things that could go wrong.I say this because csx runs through my home town.And I also agree with you about this.

One key point here is that none of these systems will ever be secure enough to be fully safe. This means that if a hacker decides to hack your car they can. Although it may take the hacker a while to do that it is still possible. Once the hacker does get control you are at their mercy, they can choose to crash your car and injure you or your family members. What is worse is that once companies feel like Machine Learning has learned enough there will be no override for a driver to use. At that point, I just can not trust a vehicle that I am not capable of controlling in case of an emergency.

Still way too many factors and variables to consider on talking autonomous trains yet.

The article from the CSX CFO also stated he mentions his age of 50,and he fills that by the end if his career he might see autonomous trains.

Maybe he retires next year or 10 years from now,but from what I've already seen it will take 10+ years to just iron out PTC.

Still way to many problems with it.

Collin "The Eastern Kentucky & Ohio R.R."

bigkid posted:
OGR Webmaster posted:
Ed Mullan posted:

I'm sure that there are a few folks who were very happy that Captain "Sully" could handle his aircraft in a very unusual situation....

Ed, I'm glad you mentioned Sully. I had not thought of that wonderful example. If this had been a pilot-less "autonomous" aircraft, everyone aboard it would be dead now, along with a lot of people on the ground.

These emergency situations are the kinds of things that technology simply cannot deal with, no matter how advanced it might be. Inevitably there will be situations presented to a vehicle (truck car, train, aircraft, whatever) that the software was not written to deal with.

The biggest, most perplexing problem presented to any software engineer is that of "error checking." What happens when things don't go as they should? Error checking within a software program has to anticipate EVERY POSSIBLE ERROR that could occur and have a routine written to deal with it.

Anticipating EVERY POSSIBLE ERROR in ANY software program is absolutely impossible. This is the "dirty little secret" that the technology proponents don't want to talk about.

I test software for a living, been doing it a long time, and that is true (there is famous theorem called the completeness theorem that says that there are things a program cannot do).   Bugs are a fact of life in software, and nothing is foolproof....but that also includes human beings, too. From everything I have been told about Captain Sully (or the pilot whose name I forgot who in the late 80's made an incredible landing I believe in Iowa, with a crippled plane, steering it with the engines alone), what they did was both incredible, and also something that may not have been handled by most other pilots..so let's say instead of Sully or the other pilot, you had one of the pilots who today fly the feeder airlines,whose training is nothing like Sully or the other ex military pilots..chances are if they were flying it, they would crash. I will add that Sully, and likely in the other example, they were relying on control systems that are heavily computerized/fly by wire) and they worked, too. For all the heroics of captain sully, how many things do we have like the Indonesia Airliner, or the pilots who commit suicide by driving the plane full of people into the ground, or who fly drunk, or who screw up like the KAL flight into SF? Not knocking pilots, but what I am saying is that human pilots are not perfect, far from it,  nor is every pilot going to pull off that once in a thousand  miracle like Sully did or the other guy I mentioned. Yeah, I have heard people sneer about their pc locking up or people on here saying "They can't even make a legacy engine that works right all the time", but that misses the point, that is comparing apples and oranges, they are built for very different things to different standards

Rich's post would be correct if we were talking straightforward programming, the kind of stuff you see in routine applications, like legacy command control,tv control, your microwave oven, smart phones, that kind of software would be very difficult to create something that could handle extraordinary circumstances, when Sully did what he did he wasn't relying on a rule book (I doubt very much there is one) or a checklist, he relied on many years of experience, he relied too on intuition and inference as well. Ordinary flying can be done by such a rule based system pretty easily, they have automated landing and takeoff systems, for example, that outside of extraordinary circumstances (something like Sully faced is probably something that happens a tiny fraction of the time, prob 99.98%  of flights are going to be routine or have issues anyone could handle) can today land and takeoff a plane, and likely would make for more efficient operations, but there were all kinds of issues around using them, some were still technical hurdles having to do with the ATC system being ancient at the time, others were resistance from the industry and pilot association. 

The thing is, that kind of programming, the kind of routine, even rule based system, programming is not what would likely be up there. One of the things that has made tremendous improvements is machine based learning, where literally systems learn as they go, they do what human beings do (in different ways), as they gain "more experience" the system becomes better and better. This  isn't science fiction, there are systems out there already using this kind of technology, systems that instead of programmers going in and rewriting routines, the systems can create their own rules based on what they come across as they go. This code isn't foolproof, and any such system would need a lot of testing and burn in time before they ever were allowed to 'replace' pilots. The thing about that kind of system is they can also 'learn' from the experience of other systems, over time as more and more data comes through from all the flights happening with these systems, the systems on board an aircraft 'learn', the way pilots today learn from talking to other pilots. So instead of a programmer guessing at situations, like "hmm, plane loses all 4 engines at 1000 feet over NYC, okay, turn the plane X degrees, set flaps to why", you could have a system that "knows" from the data ('experiences' ) of other systems, what to do/try, and to adjust if it isn't working, that is the big difference.

For the record, I don't think such systems are there yet, but in the not too near future they will be. My guess based on what I know about machine learning is initially, planes will have data recorders pulling in all kinds of data, that is used in training flight systems, helping them "learn", once they get to a certain point, have passed all kinds of simulator testing, etc, they will be onboard the aircraft, operational but not in control, and burn in further, even learning from what human pilots do...and eventually will go into trial use, perhaps with cargo planes with pilots on board acting as backup. I don't think it is going to replace pilots anytime soon, but for example, the military is actively developing pilotless planes, either controlled remotely like they do drones, or even flying themselves and operating in combat,for obvious reasons (losing a plane is a lot better than losing a pilot, if remotely controlled), plus a remotely controlled aircraft doesn't have to worry about a pilot in a dogfight getting hurt by high G forces. 

Like I wrote in another post, this has real world implications no one is really addressing, I think those who promote technology and automation and claim it is no big deal, that plenty of 'new jobs' will be created, aren't addressing the point, nor are those who scoff at what technology can do and claiming you can't replace a human being, the one thing history has shown is that both aren't true. It is true that the first phase of technology helped created more jobs, and the next ones did, but what we are seeing is different, a lot of jobs are disappearing and being replaced by very few others..and to me no one is facing this reality from any end of things.....

 

I am also in IT and there are two things no one is seriously talking about, net job loss as you mention and technilogical terroism. All of this automation relies on networking, and all of it is suceptible to hacking. World War 3 is going on now over the internet, everything that uses the internet is hackable and will be hacked at some time. Everything from your Apple watch to your home hvac controls to the computer in your car are targets and have been hacked by the bad guys. 

Imagine some state sponsored terroist hacking into the control system of trains loaded with toxic chemicals or passenger trains on the NEC. Finding the hackers is next to impossible, if we can't figure out who did it, we can't go after anyone. 

The future is exciting and scary, I hope we are able to figure out how to secure all of this automation before we start replacing humans.

Mike

Computer glitches have been occurring. 

Outside interference will most certainly be attempted. 

Nobody leaves anything alone.

And yes,job losses will occur in some way from or fashion its surely expexted.

And with most large industries going more and more automated, where are the good jobs gonna be ?

Sorta defeating the purpose of making and having good jobs for people to be a buying consumer base market.

If you dont have jobs with good or above average pay ,how can you expect people to buy higher end products ?

Collin "The Eastern Kentucky & Ohio R.R."

mackb4 posted:

Computer glitches have been occurring. 

Outside interference will most certainly be attempted. 

Nobody leaves anything alone.

And yes,job losses will occur in some way from or fashion its surely expexted.

And with most large industries going more and more automated, where are the good jobs gonna be ?

Sorta defeating the purpose of making and having good jobs for people to be a buying consumer base market.

If you dont have jobs with good or above average pay ,how can you expect people to buy higher end products ?

All very true, though if a system is designed well you can make it pretty resilient against bugs and such. Hacking could be a problem as well, depending on how they design the system it could be vulnerable to hacking, one of the biggest weaknesses often is employees doing something stupid, like clicking a link and unleashing a virus, and in a lot of companies computer security is something of a joke (for example, in this day and age, a lot of financial firms have account numbers, ss #'s and the like, stored in human readable form, or when transmitted it is sent unencrypted), and far too many  beancounters think it is a waste of money (not all companies are so stupid, but the credit card companies are notoriously bad, as are many banks, where I work we have a very vigilent computer security operation and all our systems are routinely tested for threats of all kinds, db weaknesses, through internet access, data in human readable form, you name it). 

As far as the jobs and what happens, that is a good question, all I will say is those who have blind faith that good paying jobs will be created to replace those lost are more than likely to be on the wrong end of that, and leave it at that, one thing history has shown is that a lot of companies don't understand how supply and demand work versus wages (for fun reading, look up what industrialists thought of Ford raising wages to 5 bucks a day in the 1920's). 

The person who dies with the best toys dies a happy person

bigkid posted:
mackb4 posted:

Computer glitches have been occurring. 

Outside interference will most certainly be attempted. 

Nobody leaves anything alone.

And yes,job losses will occur in some way from or fashion its surely expexted.

And with most large industries going more and more automated, where are the good jobs gonna be ?

Sorta defeating the purpose of making and having good jobs for people to be a buying consumer base market.

If you dont have jobs with good or above average pay ,how can you expect people to buy higher end products ?

All very true, though if a system is designed well you can make it pretty resilient against bugs and such. Hacking could be a problem as well, depending on how they design the system it could be vulnerable to hacking, one of the biggest weaknesses often is employees doing something stupid, like clicking a link and unleashing a virus, and in a lot of companies computer security is something of a joke (for example, in this day and age, a lot of financial firms have account numbers, ss #'s and the like, stored in human readable form, or when transmitted it is sent unencrypted), and far too many  beancounters think it is a waste of money (not all companies are so stupid, but the credit card companies are notoriously bad, as are many banks, where I work we have a very vigilent computer security operation and all our systems are routinely tested for threats of all kinds, db weaknesses, through internet access, data in human readable form, you name it). 

As far as the jobs and what happens, that is a good question, all I will say is those who have blind faith that good paying jobs will be created to replace those lost are more than likely to be on the wrong end of that, and leave it at that, one thing history has shown is that a lot of companies don't understand how supply and demand work versus wages (for fun reading, look up what industrialists thought of Ford raising wages to 5 bucks a day in the 1920's). 

Bigkid,

There is no such thing as secure software, even systems off line can be hacked. The general public only knows a very small percentage of the companies that have been hacked. We have found the Chinese in our water supply systems, and I am sure we are in theirs. 

The more we rely on systems that use the internet (99% of the worlds largest companies), the greater the risk for catastrophic terrorism. We live in scary times.

Getting back on topic, assuming the railroads are aware of the risks having an engineer put the train into an autopilot mode with the ability to turn it off and use manual controls will probably work. 

Regarding jobs of the future, there will be a lot of jobs available for people that can design, build, and run these systems as well as IT security. The skilled trades are in critical need of help and the demand will only grow as the baby boomers retire. We are going through a revolution as dramatic as the industrial revolution was if not more so. Hopefully all of us can roll with it and survive.

Mike

Add Reply

Likes (1)
SPSF


OGR Publishing, Inc. PO Box 218, Hilliard, OH 43026 330-757-3020
www.ogaugerr.com
×
×
×
×
×