Skip to main content ↑ ↓
Topic: Real Life traffic Simulators - a use for all the carefully honed sim skills (Read 5913 times) previous topic - next topic

Real Life traffic Simulators - a use for all the carefully honed sim skills

Random thoughts while driving home today.... (forgive me if this is not original)

I am temporarily living in Shanghai, where they are ripping up every road in the city in preparation for the World Expo next year. It is amazing what infinite infrastructure stimulus money can do - almost every road is being upgraded. Only 2 lanes each way? Upgrade to 3. Already have 3? Then it gets elevated. Through town traffic a little heavy? Have a new ring-road!

The final result is that I spend a lot of time sitting in construction traffic, with ample opportunity to inspect the new junctions as they are being created. Since this is the land of bicycles, mopeds, scooters, slow trucks, a million buses, plenty of cars and insane pedestrians every junction is a traffic nightmare. Older junctions are a mess - everyone just randomly shoots across with utter disregard for life and limb. The upgrades are better, but really do not address all the different forms of traffic that will be using them.

Which led me to wonder how this is done by city planners. Do they just sketch this out on a piece of paper and give it a whirl? Do they measure the traffic levels first, and then calculate the requirements? Do they have a traffic simulation program that they feed all this data into to try different options?

The reason I wondered was that I have recently been playing a LOT of Simutrans (amazing game - a big thanks to the developers), which gives me the immense satisfaction of creating complex working systems. However, other than the enjoyment factor, it is a complete waste of time. If I spent the same amount of time (and brainpower) on learning a language, I would be a polyglot (slight exaggeration, but you know what I mean...).

Would it not be wonderful if there was a global project where city planners made their "City Problems" available, with supporting traffic, map and upgrade cost data, allowing the world of simu-enthusiasts an opportunity to come up with the best solution based on safety, efficient travel flow, cost, etc etc.

Reward? Junction named after you.

Considering how much roadwork there is globally at any given time, there would be enough to keep all the Sim-fans happy for ages!

Sort of like SETI@Home, or Folding@Home, but infinitely more engaging.

End of random thoughts...

Erskie

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #1 –
Quote from: erskie – on
Which led me to wonder how this is done by city planners. Do they just sketch this out on a piece of paper and give it a whirl? Do they measure the traffic levels first, and then calculate the requirements? Do they have a traffic simulation program that they feed all this data into to try different options?

This is basicly my job (although substitute Heathrow Airport/East Anglia/South Yorkshire for Shanghai)!  You wouldn't believe how much effort it takes, tons and tons of traffic data to collect and analyse (v boring), create various models of the transport network (slightly less boring), do the tests, check the model works and is accurate and fix it when it isn't (incredibly tedious), then finally, start doing tests on what you might change and what it might do.  I say might, because models are only models, not real life, and often they are not that great (but it's the best we've got, so a lot of the non-technical people who actually make the decisions about what to build, which is politics not planning, trust them far more than the planners do!).  

So basicly, lots of number crunching and tests, and after all that it just comes down to what the client/politicians wanted in the first place...

(Ok so that was slightly cynical, but you get the point ;))

Which is why I prefer simutrans - you are the transport despot who can do anything!!! Mwahahaha  ;D

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #2 –
Ah, I am suitably dis-illusioned. :)

Still sounds interesting though. How is each scenario modelled? Can you give a hint what the software looks like? How much power does it take to simulate?

I appreciate the insight - knowing that significant effort goes into planning these things will probably help me to curse less next time I am stuck in traffic.

Erskie

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #3 –
I’m a civil engineer and so did a little bit of this at uni. Although I don’t specialise in traffic analysis as a part of my career, I do work in an office where people do. I have to agree with The Hood, it can all be very tedious, lost of numbers, fairly complex, yet somewhat inaccurate to varying degrees, computer modelling, all to justify something that the decision makers wanted in the first place.

The software used to model traffic, as with most engineering things, uses computers the way they were intended to be used, repetitive, complex, mathematical calculations. There is very little in the way of graphical representation. It’s a little like “The Matrix” in that there are just numbers on the screen, but it means so much more to someone who looks at them all day long than the average Joe. So, because there are no graphics running all over the screen the computer power required is fairly minimal. The most complex modelling I do (for flooding, hydraulics, water quality etc.) for a rather complex network will take a maximum of 30mins once I hit RUN, and that's because I have an aging computer at work. I usually use the time to make myself a coffee.

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #4 –
30 minutes isn't bad.  We have plenty of transport models that take > 1 day to converge (I think the largest one we regualarly use takes 3 days to run).  It takes so long because you try and work out how everyone gets from where they want to go from to where they want to go to on an uncongested network, then iterate their choices to take account of the congestion arising from everyone else trying to use that road/line.  If you're being smart, you also do some iterations on the number of people wanting to travel to take account of congestion effects too, and then re-do the first set of iterations all over again each time. 

Generally the best you get in terms of graphics is a stick network of the roads / transit lines with numbers/bars showing how many people are using them in an hour.  Sometimes you can do "simulated" models of small areas, but they take so much computer power it really is limited to just a few junctions at a time...

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #5 –
i'm in tianjin right now, since the subway is only under construction at the moment, congestion is incredible. the bus system cant handle the amount of p****engers and the increased number of busses clog every road. the insane taxi drivers make things worse.

there are lots of new modern junctions, built for last years olympic games. but they don't work that well. roundabouts are a bit better. but the only junction that really works must have been a roundabout once, the markings are barely noticeable now. today it's a criss-cross-wherever-you-want-whenever-you want. and it really works. No signaling, no markings, obstacles or anything. trafic just flows chaoticaly through it. it seems planing made just things worse, self orgnisation works well however. (this is very very true for most aspects of chinese organisation)

police picks crossings at random to coordinate traffic. this requires about 4 real policemen and a dozen helpers blocking the cycle lanes. They are very indoustrious and effective. everything is very orderly, and the traffic jam they cause reaches three junctions back.


@the hood: physicist?

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #6 –
Wow, models taking greater than three days to run, that’s insane. The only ones I have heard of taking that order of time is for structural design iterations, such as for the Water Cube in Beijing.

I hate to think of the number of mathematical equations undertaken in that time. I think it has often been said, no matter how good computers get, engineers will very quickly use them to full capacity.

If you don’t mind me asking, how long does it take to set up a model like that in the first place, before even thinking about running it?
 

Re: Real Life traffic Simulators - a use for all the carefully honed sim skills

Reply #7 –
Quote from: The Hood – on
30 minutes isn't bad.  We have plenty of transport models that take > 1 day to converge (I think the largest one we regualarly use takes 3 days to run).  It takes so long because you try and work out how everyone gets from where they want to go from to where they want to go to on an uncongested network, then iterate their choices to take account of the congestion arising from everyone else trying to use that road/line.  If you're being smart, you also do some iterations on the number of people wanting to travel to take account of congestion effects too, and then re-do the first set of iterations all over again each time. 

Generally the best you get in terms of graphics is a stick network of the roads / transit lines with numbers/bars showing how many people are using them in an hour.  Sometimes you can do "simulated" models of small areas, but they take so much computer power it really is limited to just a few junctions at a time...

For rail stuff, I've seen some graphics in professional use which *I* think are rather pretty, namely the full dispatching models; line diagrams such as dispatchers use, with contrast-lighted sections indicating trains moving.  Simple, highly abstract, but effective.  These appear to be used to squeeze the maximum productivity out of an expensive, long rail line, and to test whether small bottleneck fixes can allow even more trains to be squeezed in.  Only seen this in articles, my job has nothing to do with transport.  Apparently these are really computer-intensive models., but also quite accurate  But I suspect they're not that hard to build, because a lot of the programming is very similar to what has to be done for actual dispatching software (which is difficult and expensive, but is also *already done*).  I've never heard of this level of detail modeling outside rail.  It's a lot more detailed than simutrans's model.  :-)

---

For roads, there rather seems to be a principle that traffic control is only really valuable if you have traffic faster than a certain speed; below a certain point self-organization at the intersection works very well.  Once you're going faster than people's ability to see what's going on and react, *then* you need traffic control.  For trains this principle is somewhat less true, because the "critical speed" is the speed at which the brakeman can jump out, change the switch, wait for the train to go by, change the switch, and jump back on, which is rather lower than the speed at which a car driver can react.

Of course, this ****umes a certain level of intelligence on the part of "people".  The traffic light was introduced after a rather famous accident in Chicago in which two cars, both driving down the middle of the road, crashed head-on into each other.  Horses were never stupid enough to do that, so traffic lights were not necessary until motorcars, and directional lanes were rarely needed either!