1897: "Self Driving"

This forum is for the individual discussion thread that goes with each new comic.

Moderators: Moderators General, Prelates, Magistrates

Mikeski
Posts: 876
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1897: "Self Driving"

Postby Mikeski » Wed Oct 04, 2017 12:48 am UTC

It's also harder than the traveling salesman problem since, with today's tech, the passengers probably expect the system to be intelligent w.r.t. traffic/construction/accidents/etc, choosing a route that minimizes time at the possible expense of distance. So not only is the list of destinations changing, but some legs of the journey may change "distance" dynamically, as well.

It also starts with the asymmetric traveling-salesman problem (the distance from A to B is not necessarily the same as the distance from B to A), since buses can't go the wrong way down one-way streets... even though they get to break certain other traffic laws, at least in Minneapolis.

I suppose there's something that can be done in precomputing part of the solution space (take as much time as you want and brute-force it), because a "bus" only stops in certain predefined places. I assume I can't ask the Minneapolis bus to take me to Chicago, or drop me off in the middle of the Lowry Hill Tunnel, even if I'm the only person on board.

Mikeski
Posts: 876
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1897: "Self Driving"

Postby Mikeski » Wed Oct 04, 2017 1:01 am UTC

jc wrote:Shipping companies might run into a version of this a lot, where customers are constantly sending in new requests, while the fleet's drivers are driving along whatever routes the company controllers last sent them, and they're probably used to getting route changes at many stops.


Actually, that's unlikely. My brother works for a very large company (their market cap is similar to UPS's), and they couldn't get a UPS truck to change routes. They needed a shipment by X o'clock (due to the guy needing it leaving on an airplane), and had "guaranteed" delivery by X-minus-Y o'clock, and didn't get it.

Panicked phone call, and they found out it wouldn't make it by X o'clock, either.

The best UPS would do was give them the truck's route, so one of his co-workers could intercept the truck at a different stop. (And refund their "guaranteed" delivery fee.)

User avatar
roderik
Posts: 22
Joined: Fri Apr 10, 2009 4:16 pm UTC
Location: Below sealevel

Re: 1897: "Self Driving"

Postby roderik » Wed Oct 04, 2017 1:07 am UTC

jonhaug wrote:
cellocgw wrote:. But some clever hacker-type guys have figured out how to change a few pixels in the image and get the algo to completely fail -- while not changing the image in a way noticeable to the human Mark-I eyeball.


Well, this is just a typical game theoretical arms race.

(1) You solve a problem.
(2) Someone breaks it.
(3) You fix
(4) Goto (2)

Things do get better over time.

IIRC there's even some types of learning algorithms that use this sort of thing, where one part is constantly trying to find ways to cheat the system, and another tries to figure out what was real data or fake data.

sonar1313
Posts: 85
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1897: "Self Driving"

Postby sonar1313 » Wed Oct 04, 2017 1:42 am UTC

orthogon wrote:
Heimhenge wrote:OP Envelope Generator said: "Crowdsourced steering" makes me think "Let Twitch drive you to work".

Reminded me of a serious (?) paper I read some time ago about "democratic self-driving buses" where the all passengers would have keypads for entering their desired destination. The bus then processes that data and computes the optimal route for its current riders. Wasn't so much "AI" as it was solving for the optimal path between nodes in a network. It might have been a paper on game theory.

Made sense to me at the time. Then I forgot about it until this comic. Wondering now if something like that is part of the AI in multi-passenger autonomous vehicles (like taxis).

Isn't that exactly the Traveling Salesman problem? I'd heard of it mainly as a computational complexity thing, but I guess AI approaches could find good suboptimal solutions.

Nah, because the bus has to stop at bus stops (so that others will be able to get on) and theoretically (I have not read this paper) might not even stop at the best bus stop for a particular destination, figuring that one guy walking an extra five minutes is worth shaving ten off the route for the rest of the riders. Or it skips a stop for the same reason - maybe five people want one stop and one wants another, and it's more efficient to make the guy walk to his stop from the one that the five want. In short, Traveling Salesman requires touching every stop and weights them all the same; such a bus would not.

Anyway, I have a hard time believing such a "democratic self-driving bus" could ever work, because a really really important part of the point of a bus is that it shows up in the same place, at the same time, every day.

sonar1313
Posts: 85
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1897: "Self Driving"

Postby sonar1313 » Wed Oct 04, 2017 1:49 am UTC

Mikeski wrote:
jc wrote:Shipping companies might run into a version of this a lot, where customers are constantly sending in new requests, while the fleet's drivers are driving along whatever routes the company controllers last sent them, and they're probably used to getting route changes at many stops.


Actually, that's unlikely. My brother works for a very large company (their market cap is similar to UPS's), and they couldn't get a UPS truck to change routes. They needed a shipment by X o'clock (due to the guy needing it leaving on an airplane), and had "guaranteed" delivery by X-minus-Y o'clock, and didn't get it.

Panicked phone call, and they found out it wouldn't make it by X o'clock, either.

The best UPS would do was give them the truck's route, so one of his co-workers could intercept the truck at a different stop. (And refund their "guaranteed" delivery fee.)

I agree it's unlikely, for the reason that the delivery routes are optimized to save gas, not time. Constantly changing them would crank up gas costs pretty heftily as trucks zigzagged around town.

User avatar
Flumble
Yes Man
Posts: 1951
Joined: Sun Aug 05, 2012 9:35 pm UTC

Re: 1897: "Self Driving"

Postby Flumble » Wed Oct 04, 2017 10:48 am UTC

sonar1313 wrote:Anyway, I have a hard time believing such a "democratic self-driving bus" could ever work, because a really really important part of the point of a bus is that it shows up in the same place, at the same time, every day.

That's how it's been for buses (and other forms of public transport), but it's not an inherent property.
Though, even if nearly all passengers would tell the system when and where they want to go a day in advance, it may not be more efficient (in terms of waiting time, travel time, energy consumed, strain on traffic) than the current situation. Buses would go all over the place and increase traffic (because of their probably independent routes* and random timing), people will still have to wait for the bus and may not have the shortest route to their destination and those few (or a lot in the real world) unplanned passengers will screw over any efficiency statistic.


*on a tangent, that reminds me of Braess' paradox

rmsgrey
Posts: 3106
Joined: Wed Nov 16, 2011 6:35 pm UTC

Re: 1897: "Self Driving"

Postby rmsgrey » Wed Oct 04, 2017 11:23 am UTC

Around here, you can divide buses into two general categories:

A) Low frequency, high reliability buses that (with occasional exceptions due to mechanical trouble) follow the advertised times to within a minute or so
B) High frequency, low reliability buses that run, say, 8 buses an hour, but the timing of any given bus arriving at the stop has no correlation with the timetable

Type A buses, you plan for; type B buses, you turn up at the stop and wait for the next one to turn up - usually with less than 10 minutes wait time.

sonar1313
Posts: 85
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1897: "Self Driving"

Postby sonar1313 » Wed Oct 04, 2017 1:27 pm UTC

Flumble wrote:
sonar1313 wrote:Anyway, I have a hard time believing such a "democratic self-driving bus" could ever work, because a really really important part of the point of a bus is that it shows up in the same place, at the same time, every day.

That's how it's been for buses (and other forms of public transport), but it's not an inherent property.
Though, even if nearly all passengers would tell the system when and where they want to go a day in advance, it may not be more efficient (in terms of waiting time, travel time, energy consumed, strain on traffic) than the current situation. Buses would go all over the place and increase traffic (because of their probably independent routes* and random timing), people will still have to wait for the bus and may not have the shortest route to their destination and those few (or a lot in the real world) unplanned passengers will screw over any efficiency statistic.


*on a tangent, that reminds me of Braess' paradox


Well, expanding on the purpose of a bus, it actually is an inherent property. Public transportation has to be relied upon to show up at the expected time and place. Otherwise you have to schedule it for pickup, in which case it becomes more of a vanpool or a shared Uber than a bus. And part of the point of public transportation is that you don't need to schedule it; in a pinch, you can pay the fare and jump aboard. A bus whose routes are always changing and must be scheduled ceases to be a bus as we know it, and doesn't replace the need for a bus as we know it.

User avatar
orthogon
Posts: 2720
Joined: Thu May 17, 2012 7:52 am UTC
Location: The Airy 1830 ellipsoid

Re: 1897: "Self Driving"

Postby orthogon » Wed Oct 04, 2017 3:16 pm UTC

sonar1313 wrote:
Flumble wrote:
sonar1313 wrote:Anyway, I have a hard time believing such a "democratic self-driving bus" could ever work, because a really really important part of the point of a bus is that it shows up in the same place, at the same time, every day.

That's how it's been for buses (and other forms of public transport), but it's not an inherent property.
Though, even if nearly all passengers would tell the system when and where they want to go a day in advance, it may not be more efficient (in terms of waiting time, travel time, energy consumed, strain on traffic) than the current situation. Buses would go all over the place and increase traffic (because of their probably independent routes* and random timing), people will still have to wait for the bus and may not have the shortest route to their destination and those few (or a lot in the real world) unplanned passengers will screw over any efficiency statistic.


*on a tangent, that reminds me of Braess' paradox


Well, expanding on the purpose of a bus, it actually is an inherent property. Public transportation has to be relied upon to show up at the expected time and place. Otherwise you have to schedule it for pickup, in which case it becomes more of a vanpool or a shared Uber than a bus. And part of the point of public transportation is that you don't need to schedule it; in a pinch, you can pay the fare and jump aboard. A bus whose routes are always changing and must be scheduled ceases to be a bus as we know it, and doesn't replace the need for a bus as we know it.


I had the same feeling as Flumble, that there were quite a few features of what a bus is and isn't, none of which should necessarily be taken for granted in the definition of a future bus. Already buses in London are a qualitatively different thing to what they used to be, thanks to technology: whilst rmsgrey is essentially right...

rmsgrey wrote:Around here, you can divide buses into two general categories:

A) Low frequency, high reliability buses that (with occasional exceptions due to mechanical trouble) follow the advertised times to within a minute or so
B) High frequency, low reliability buses that run, say, 8 buses an hour, but the timing of any given bus arriving at the stop has no correlation with the timetable

Type A buses, you plan for; type B buses, you turn up at the stop and wait for the next one to turn up - usually with less than 10 minutes wait time.


... the availability of live bus arrival information on your smartphone introduces a kind of intermediate ("AB") category. You can surf up the bus times and head to the stop when the bus is about to arrive, which means that relatively infrequent buses with almost-Poisson arrivals can still be very useful.

I live near to a sort of special bus route (the G1, if anyone knows it): the bus goes literally all around the houses and takes forever but provides people who aren't very mobile and can afford the time with a great service. On parts of its route there are no stops: it just stops as soon as it's safe after somebody presses the button, and you can hail it from anywhere along that stretch. The route is kind of an approximation to a fractal space-filling curve; I exaggerate, but it often passes very close to where it's been before or goes back on itself. This kind of route would work much better if the bus knew where its passengers were and where they wanted to go, since it could miss out whole sections of its route.
xtifr wrote:... and orthogon merely sounds undecided.

jgh
Posts: 91
Joined: Thu Feb 03, 2011 1:04 pm UTC

Re: 1897: "Self Driving"

Postby jgh » Wed Oct 04, 2017 4:10 pm UTC

Soupspoon wrote:For the past few days, now, I've been trying to ID a comedy sketch (could be anything from Monty Python to Little & Large, as far as my memory goes for what show it was on) which had a bus where every seat had a steering wheel (and implied accelerator/brake?), so that everybody had an input. Perhaps according to how much they fed into their personal coin-slot, to dictate how much proportional influence everybody input to the vehicle.

Daedalus, New Scientist's DREADCO correspondant. Some of his columns (including the crowd-steered bus) were made into a TV show.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 2528
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1897: "Self Driving"

Postby Soupspoon » Wed Oct 04, 2017 4:54 pm UTC

jgh wrote:Daedalus, New Scientist's DREADCO correspondant. Some of his columns (including the crowd-steered bus) were made into a TV show.
It may have been done there too, but this was an established general comedy show. I seem to have compleyely missed any NS crossover onto TV, or any mention in NS for the decade or so I was once personally subscribed.

(I may link it to Little And Large by memories of a sketch of theirs that had Eddie Large roll up to Sid Little in a ?Golf Club? car-park doing a side-wheelie/two-wheel/ski mode thing, before stopping and dropping back to level. Asked how long he'd been driving around like that, he replies "ever since I got this!", pointing at a wheel-clamp on the previously off-ground front wheel. OTOH, I also get visions (probably erroneously) of the bus being populated by a load of Pepperpots... Either way, I can't seem to find any Youtube uploads of it, so far, so it might as well just be my own perverse imagination.)

User avatar
Archgeek
Posts: 128
Joined: Wed May 02, 2007 6:00 am UTC
Location: Central US
Contact:

Re: 1897: "Self Driving"

Postby Archgeek » Wed Oct 04, 2017 5:50 pm UTC

speising wrote:
Weeks wrote:
jonhaug wrote:I frequently fail at the Türing test when I typically get "Click on the images that contain a car" and have to try over and over.

/Jon
The Türing test, named after Ålan Türing

Türiñ.

Well that just sounds like a test to find out if someone's an actual Dwarf and not a human with dwarfism -- the Durin test.
"That big tube down the side was officially called a "systems tunnel", which is aerospace contractor speak for "big tube down the side."

sonar1313
Posts: 85
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1897: "Self Driving"

Postby sonar1313 » Wed Oct 04, 2017 6:11 pm UTC

orthogon wrote:
sonar1313 wrote:
Flumble wrote:
sonar1313 wrote:Anyway, I have a hard time believing such a "democratic self-driving bus" could ever work, because a really really important part of the point of a bus is that it shows up in the same place, at the same time, every day.

That's how it's been for buses (and other forms of public transport), but it's not an inherent property.
Though, even if nearly all passengers would tell the system when and where they want to go a day in advance, it may not be more efficient (in terms of waiting time, travel time, energy consumed, strain on traffic) than the current situation. Buses would go all over the place and increase traffic (because of their probably independent routes* and random timing), people will still have to wait for the bus and may not have the shortest route to their destination and those few (or a lot in the real world) unplanned passengers will screw over any efficiency statistic.


*on a tangent, that reminds me of Braess' paradox


Well, expanding on the purpose of a bus, it actually is an inherent property. Public transportation has to be relied upon to show up at the expected time and place. Otherwise you have to schedule it for pickup, in which case it becomes more of a vanpool or a shared Uber than a bus. And part of the point of public transportation is that you don't need to schedule it; in a pinch, you can pay the fare and jump aboard. A bus whose routes are always changing and must be scheduled ceases to be a bus as we know it, and doesn't replace the need for a bus as we know it.


I had the same feeling as Flumble, that there were quite a few features of what a bus is and isn't, none of which should necessarily be taken for granted in the definition of a future bus. Already buses in London are a qualitatively different thing to what they used to be, thanks to technology: whilst rmsgrey is essentially right...

rmsgrey wrote:Around here, you can divide buses into two general categories:

A) Low frequency, high reliability buses that (with occasional exceptions due to mechanical trouble) follow the advertised times to within a minute or so
B) High frequency, low reliability buses that run, say, 8 buses an hour, but the timing of any given bus arriving at the stop has no correlation with the timetable

Type A buses, you plan for; type B buses, you turn up at the stop and wait for the next one to turn up - usually with less than 10 minutes wait time.


... the availability of live bus arrival information on your smartphone introduces a kind of intermediate ("AB") category. You can surf up the bus times and head to the stop when the bus is about to arrive, which means that relatively infrequent buses with almost-Poisson arrivals can still be very useful.


Possibly I'm not being clear. If a route-optimization system is in place that changes based on who's on the bus, the bus is going to never appear at some stops. Not drive right past them - it simply will never be there. So you call up the current route on your fancy app and you discover that because of the optimized route for today, you have a twenty-minute walk to catch the earliest possible bus, because the nearest bus stop got optimized right out of the route.

Buses already have route optimization in a sense, because they have a driver who can skip a stop if there's nobody waiting and nobody has pulled the string. Buses generally have primary stops and secondary stops - they will always stop at the primaries, but there are other locations you can get on and off that they'll skip if nobody asked for it. The driver can make that decision; self-driving buses where the passengers have to do something more than pull a string complicate the issue rather than simplify it. Of course, the route itself is fixed. And that's what really matters. People need to know they can sit at a bus stop and the bus will be there - not have to look up whether that stop was on today's route or not.

You could have a service where people schedule a pickup and dropoff, a shared Uber of sorts, but as I mentioned before, that doesn't replace the need for a service that people can rely on to appear at a set place and time.

User avatar
roderik
Posts: 22
Joined: Fri Apr 10, 2009 4:16 pm UTC
Location: Below sealevel

Re: 1897: "Self Driving"

Postby roderik » Thu Oct 05, 2017 2:07 am UTC

sonar1313 wrote:Possibly I'm not being clear. If a route-optimization system is in place that changes based on who's on the bus, the bus is going to never appear at some stops. Not drive right past them - it simply will never be there. So you call up the current route on your fancy app and you discover that because of the optimized route for today, you have a twenty-minute walk to catch the earliest possible bus, because the nearest bus stop got optimized right out of the route.

Buses already have route optimization in a sense, because they have a driver who can skip a stop if there's nobody waiting and nobody has pulled the string. Buses generally have primary stops and secondary stops - they will always stop at the primaries, but there are other locations you can get on and off that they'll skip if nobody asked for it. The driver can make that decision; self-driving buses where the passengers have to do something more than pull a string complicate the issue rather than simplify it. Of course, the route itself is fixed. And that's what really matters. People need to know they can sit at a bus stop and the bus will be there - not have to look up whether that stop was on today's route or not.

You could have a service where people schedule a pickup and dropoff, a shared Uber of sorts, but as I mentioned before, that doesn't replace the need for a service that people can rely on to appear at a set place and time.



Another issue with such a system is that you can't count on transfers to other buses(or trains or planes or whatever) to be in time. (at least without yet more algorithmic complexity)

What use is there if your bus optimises the path to your next stop, but you end up waiting half an hour there anyway waiting on your next bus.

elasto
Posts: 3125
Joined: Mon May 10, 2010 1:53 am UTC

Re: 1897: "Self Driving"

Postby elasto » Thu Oct 05, 2017 4:48 pm UTC

You could have a service where people schedule a pickup and dropoff, a shared Uber of sorts, but as I mentioned before, that doesn't replace the need for a service that people can rely on to appear at a set place and time.


What use is there if your bus optimises the path to your next stop, but you end up waiting half an hour there anyway waiting on your next bus.


I dunno how cheap busses are for you guys, but they are pretty expensive in the UK. I could easily see driverless uber cars becoming as cheap as busses are today. In a couple of decades probably only the truly frugal will need to catch multiple busses as part of a single journey, just like how today it's perfectly economical to hop on a plane rather than need take a cross-country coach.

rmsgrey
Posts: 3106
Joined: Wed Nov 16, 2011 6:35 pm UTC

Re: 1897: "Self Driving"

Postby rmsgrey » Fri Oct 06, 2017 11:00 am UTC

elasto wrote:
You could have a service where people schedule a pickup and dropoff, a shared Uber of sorts, but as I mentioned before, that doesn't replace the need for a service that people can rely on to appear at a set place and time.


What use is there if your bus optimises the path to your next stop, but you end up waiting half an hour there anyway waiting on your next bus.


I dunno how cheap busses are for you guys, but they are pretty expensive in the UK. I could easily see driverless uber cars becoming as cheap as busses are today. In a couple of decades probably only the truly frugal will need to catch multiple busses as part of a single journey, just like how today it's perfectly economical to hop on a plane rather than need take a cross-country coach.


Last time I checked (though it has been a while), buses were still cheaper than owning and maintaining a car for most purposes - between fuel, road tax, MOT, actual repairs, and the depreciation of the car's value over time, the actual cost of running a car is a lot more than most people realise.

Similarly, for most people, a plane journey costs more and takes longer than you realise - airports are generally some distance out of town, so there's the cost of getting there and getting from the destination airport to the actual destination. A couple of years ago, I sat down and worked out how long it would take, and how much it would cost, door-to-door, to get from where I live (Tyne and Wear) to my parents' house (Greater London) and it worked out that the cheapest was to go by coach, and the fastest by train - plane didn't have the advantage in either category.

User avatar
Steve the Pocket
Posts: 646
Joined: Mon Apr 23, 2007 4:02 am UTC
Location: Going downtuuu in a Luleelurah!

Re: 1897: "Self Driving"

Postby Steve the Pocket » Sun Oct 08, 2017 12:22 am UTC

orthogon wrote:(I have a counter-theory: my incisors play the role of the snare drum in my internal drumkit with which I accompany my humming. Anybody else do this?)

I have too severe of an overbite for that, so I use my back teeth for all of it. Bite down for kick/tom, grind laterally for snare. I can do both simultaneously (e.g. for cases where a kick on every beat plus snare on every other is appropriate) by biting down on the left and quickly snapping my jaw to the right. And I can get a drumroll going by rapidly vibrating my jaw side to side so the bottom row repeatedly slams into the insides of the uppers.

Somewhere a dentist is crying.

(It's weird. I've been doing this all my life and only now have I brought it up to anyone.)
cephalopod9 wrote:Only on Xkcd can you start a topic involving Hitler and people spend the better part of half a dozen pages arguing about the quality of Operating Systems.

Baige.

User avatar
orthogon
Posts: 2720
Joined: Thu May 17, 2012 7:52 am UTC
Location: The Airy 1830 ellipsoid

Re: 1897: "Self Driving"

Postby orthogon » Sun Oct 08, 2017 12:33 pm UTC

Steve the Pocket wrote:
orthogon wrote:(I have a counter-theory: my incisors play the role of the snare drum in my internal drumkit with which I accompany my humming. Anybody else do this?)

I have too severe of an overbite for that, so I use my back teeth for all of it. Bite down for kick/tom, grind laterally for snare. I can do both simultaneously (e.g. for cases where a kick on every beat plus snare on every other is appropriate) by biting down on the left and quickly snapping my jaw to the right. And I can get a drumroll going by rapidly vibrating my jaw side to side so the bottom row repeatedly slams into the insides of the uppers.

Somewhere a dentist is crying.

(It's weird. I've been doing this all my life and only now have I brought it up to anyone.)

I'm so pleased I'm not alone. (To be honest, once I started thinking about it, I couldn't be sure which teeth I naturally used for which drum - I'd have to "catch myself" doing it). Like you, I've never mentioned it to anyone before; it's one of those things that you sort of assume everybody does until you think about how private it is. Have you looked at yourself in the mirror doing it? The muscle movement is just about discernible, and the lips move a bit, but I'm not sure that an observer would notice, still less identify what was going on. The great thing is how loud it is internally, owing to bone conduction.
xtifr wrote:... and orthogon merely sounds undecided.


Return to “Individual XKCD Comic Threads”

Who is online

Users browsing this forum: No registered users and 35 guests