Page 1 of 1
| Getting the best out of ARS 03/09/2020 at 14:18 #131264 | |
Afterbrunel
94 posts |
It would be helpful if a professional would write up some notes about getting the best out of ARS. How much to intervene? How best to stop unwanted routes being set up (when everything is out of course).
Log in to reply The following user said thank you: ambergatesm |
| Getting the best out of ARS 04/09/2020 at 03:19 #131277 | |
|
Chromatix
190 posts |
Not a real-life professional in this area, but this is what I do in sim: 1: Reminders are your friend with ARS, as ARS won't set a route to or from a signal with a reminder set on it. Other ways of inhibiting ARS operation are to set individual trains non-ARS, or pull (right-click) the purple roundel controlling an ARS area. It takes a fair amount of experience to learn precisely which signals each area corresponds to, though. In some cases I have also manually keyed points (left-click for normal, right-click for reverse) to prevent ARS routes in one direction but permit them in another, or to direct them over a particular alternative among several possible routes (which is possible at some station throats). 2: ARS usually tries to route trains over a conflict in timetabled order. Some junctions may have a different rule implemented (particularly first-come first-served), or limit the area in which trains are searched for regarding the timetabled-order rule, or even something more complex which takes train classes into account. It's often not obvious what ARS will do next during disruption, so if it really matters for regulation and is out of the usual order, inhibit ARS as above and set the routes manually. 3: You can often unstick ARS from circular dependencies or waiting for a severely delayed train by simply setting a route manually for a waiting train, without manually inhibiting anything. Beware when doing so that you are setting the correct route for that train - it's easy to fall into habits, that then lead you astray when the service pattern changes. In particular, check which direction the next working departs in before manually setting a route into a bay platform or one occupied by another service, unless the bay is what is timetabled. In most other cases the driver will phone you to complain about a wrong-routing, but that's actually preferable to having to walk a train through a shunt move! 4: During disruption it is often helpful to divert trains off their booked line to a parallel one, or into a loop. This will automatically mark the train as Non-ARS. Once the train is established on the alternative path (ie. the TD is actually on it rather than just approaching), you can make the train ARS again and it will try to find its way back to the booked path. You might reasonably preset the route for an overtaking train before doing so, forcing ARS to let the diverted train out only behind it. There are some parallel paths, however, for which you can't re-enable ARS until the train has mostly completed it - eg. swapping between the station and the avoiding lines at York, or when a train enters on the "wrong" line from Wimbledon on the Waterloo sim. 5: When a disruption event effectively closes a line, such that you have to route all trains down one line instead of two, I sometimes set a reminder on the signals entering both lines, then use Reminder Override to manually set routes. This is particularly useful when traffic is heavy enough on the open line to make it hard to squeeze an extra train in before ARS automatically sets a route for the next one. You might also need to do the same for the opposite-direction line if you need to cross it. 6: You can get some insight into what ARS is "thinking" by observing the ARS status in the timetable popup. Note that this is only valid for the moment the popup content was updated, and can get very stale if you leave it open for a while. Log in to reply The following users said thank you: whatlep, Afterbrunel |
| Getting the best out of ARS 04/09/2020 at 08:46 #131279 | |
Hap
1039 posts |
Automatic Route Setting: information is sent to three computers, two of which have to agree an outcome before the route can automatically be set for the train. You can find some info on ARS in the below link. https://www.SimSig.co.uk/Wiki/Show?page=usertrack:ars Just as a point of order, train classes...Class 1s won't necessarily be prioritised over a class 2. Although ARS is there to do a job of route setting, it is there as an assistant and not to take over full control of the railway, manual intervention will be required to regulate (especially trains that aren't ARS able at all. but getting into the habit of thinking that because a class 2 is apparently holding up a class 1 then you must find a way to get the class 1 in front. By doing that you may actually snooker yourself further down the line with another train or even get a train join or detach in the wrong order. For any train taken off route, it will not go back to its ARS state until it has been routed back onto it's booked route. In times of disruption, you aren't going to get the best out of ARS, even in real life. The signaller will have to ensure that trains are running in some sort of booked order, again for the reasons above. Trains are on booked diagrams, booked on set routes etc etc. It really depends how you intend to sim..(As close to the real thing or just for fun) as to how you handle it. How to report an issue: www.SimSig.co.uk/Wiki/Show?page=usertrack:reportanissue Last edited: 04/09/2020 at 08:48 by Hap Reason: None given Log in to reply The following user said thank you: whatlep |
| Getting the best out of ARS 04/09/2020 at 10:45 #131281 | |
|
clive
2789 posts |
Hap in post 131279 said:Automatic Route Setting: information is sent to three computers, two of which have to agree an outcome before the route can automatically be set for the train.That's a standard mechanism for reliability in safety-critical situations. SSIs have three separate processors which constantly compare their internal states. If two disagree with the third, that third one is turned off (by blowing a fuse). If the two that are left disagree, they both turn off. Log in to reply |
| Getting the best out of ARS 04/09/2020 at 15:56 #131287 | |
|
Stephen Fulcher
2080 posts |
Had a good fault at Leamington once where one of the Interlocking MPMs kept blowing a security fuse and was worked out in the end as one of the other two being faulty causing it.
Log in to reply |
| Getting the best out of ARS 04/09/2020 at 16:16 #131291 | |
GeoffM
6376 posts |
A good write up, however, a couple of minor points: Chromatix in post 131277 said: 2: ARS usually tries to route trains over a conflict in timetabled order. Some junctions may have a different rule implemented (particularly first-come first-served), or limit the area in which trains are searched for regarding the timetabled-order rule, or even something more complex which takes train classes into account.No part of ARS has first-come-first-served: it's the same strategy everywhere. Every train being considered for routesetting checks itself against every other train visible to ARS, no matter where it is in the area (which means if it's not visible, then ARS does not consider it). A complicated set of mathematics and predictions brings it down to a score for every pair of trains. And yes, sometimes the number produced is not what a human would like. But it's a machine with rules and is simply following the rules set up. Hap in post 131279 said: Automatic Route Setting: information is sent to three computers, two of which have to agree an outcome before the route can automatically be set for the train.To clarify: ARS is safety-related (SIL2) and single processor (warm standby), whereas the interlockings are safety-critical (SIL4). The request from ARS to set a route is sent over the currently-active comms link, then that request gets disseminated to the three computers (SSI/Westlock/Smartlock/relay-interface). SimSig Boss Last edited: 04/09/2020 at 16:19 by GeoffM Reason: None given Log in to reply |
| Getting the best out of ARS 05/09/2020 at 10:22 #131326 | |
|
kbarber
1743 posts |
GeoffM in post 131291 said:A good write up, however, a couple of minor points:The mathematics are probably quite interesting if you're into that sort of thing. A year or so back I was following links around the web (as you do) and was reading about the BR Research setup, whose staff included mathematicians. There was a link to a paper read at one of the leading mathematical societies, the mathematics of regulating trains at a simple double junction. Seriously; that was worth reading at that level. So there's more to this signalling stuff than meets the eye. My eyes glazed over before I reached the end of the first paragraph. But the notes made it clear that this was the origin of the ARS system. Which does in a rule-bound way exactly what a good signalman learns to do intuitively as a matter of course: where there's a possible conflict, decide which train will have proceedings as late as possible, but in time to give it a clear road. Which just goes to underline what an amazing piece of kit the Mk I human brain is. Last edited: 05/09/2020 at 10:22 by kbarber Reason: None given Log in to reply The following user said thank you: whatlep |
| Getting the best out of ARS 07/09/2020 at 10:28 #131394 | |
|
Chromatix
190 posts |
To be fair, the formal analysis of even quite simple algorithms can generate an awful lot of opaque-looking mathematics. I've just been going over a 2016 paper related to Quicksort, which was invented in 1962 and whose fundamentals can be expressed in about half a dozen lines of C (though such a simple implementation is easy to drive into pathological behaviour). Shellsort (1959) is also still under active study, though less urgently, and is even easier to implement - the research all goes into the gap sequences. Quote: No part of ARS has first-come-first-served: it's the same strategy everywhere. Every train being considered for routesetting checks itself against every other train visible to ARS... I'm pretty sure I've seen behaviour that differs from that assertion in SimSig. In one case it was very clearly a sim bug, but what about the others? Log in to reply |
| Getting the best out of ARS 07/09/2020 at 15:16 #131399 | |
|
bill_gensheet
1413 posts |
Some non-ARS sims have a simple first come area for say single track branches or (maybe) sections. One of the recently fixed ARS bugs could create a first come first served effect at certain track geometries as ARS was doing 'train needs another route section' but was not picking up & checking conflicting trains. Bill Log in to reply |
| Getting the best out of ARS 07/09/2020 at 15:35 #131400 | |
|
clive
2789 posts |
kbarber in post 131326 said:I don't suppose you have a URL, do you? Log in to reply |
| Getting the best out of ARS 07/09/2020 at 17:12 #131402 | |
|
Sacro
1171 posts |
clive in post 131400 said:kbarber in post 131326 said:Might W.-S. Lin and J.-W. Sheu, “Automatic train regulation for metro lines using dual heuristic dynamic programming,” be the one?I don't suppose you have a URL, do you? Last edited: 07/09/2020 at 17:12 by Sacro Reason: None given Log in to reply |
| Getting the best out of ARS 07/09/2020 at 20:58 #131410 | |
|
Jan
906 posts |
2009 seems much too recent for ARS.
Two million people attempt to use Birmingham's magnificent rail network every year, with just over a million of them managing to get further than Smethwick. Log in to reply |
| Getting the best out of ARS 08/09/2020 at 08:53 #131417 | |
|
kbarber
1743 posts |
clive in post 131400 said:kbarber in post 131326 said:I don't suppose you have a URL, do you? Sorry Clive, I don't recall quite what I was following let alone what caught my eye. If I do come across it I'll let you know. Log in to reply |
| Getting the best out of ARS 08/09/2020 at 17:16 #131418 | |
|
GeoffM
6376 posts |
kbarber in post 131326 said:A year or so back I was following links around the web (as you do) and was reading about the BR Research setup, whose staff included mathematicians. There was a link to a paper read at one of the leading mathematical societies, the mathematics of regulating trains at a simple double junction. Seriously; that was worth reading at that level. So there's more to this signalling stuff than meets the eye.That could have been my old boss, Harry Ryland. Jan in post 131410 said: 2009 seems much too recent for ARS.No reason why the paper couldn't have been written later. I know he did write and present papers on various subjects when I worked for him (2000-2012). Our company wrote the "ARS plus" specification that Deltarail (the IECC maintainers) and others had to adopt, which (as the name suggests) was ARS with extra features like simulation, a "clear the area" regulation strategy, perturbation plans, more I don't recall. SimSig Boss Log in to reply |
| Getting the best out of ARS 08/09/2020 at 19:44 #131427 | |
|
Jan
906 posts |
True as well, though in that particular instance I still don't think it was that specific article that Keith came across. Thinking a little about it, I remember stumbling across some paper/article related to ARS as well, but can't remember what/where exactly, either. The closest I can find in my archive is an article called History of British Railways' electrical research by a certain Mr A.O. Gilchrist (it's a small world, isn't it?), which spends a few paragraphs on ARS as well, but in the end it's not the one I'm trying to think of… Two million people attempt to use Birmingham's magnificent rail network every year, with just over a million of them managing to get further than Smethwick. Log in to reply |
| Getting the best out of ARS 09/09/2020 at 09:02 #131441 | |
|
kbarber
1743 posts |
GeoffM in post 131418 said:kbarber in post 131326 said:I'm quite sure the paper I came across was 1970s or even late '60s. It was totally mathematics, apart from explaining the layout of a double junction and the possible moves and conflicts. It had been read, as I say, at a mathematical society (maybe even The Mathematical Society of the UK - though I don't know if that's its proper name), which I imagine is pretty much equivalent to being published in a peer reviewed journal. And although it (as far as I can remember) contented itself with working out mathematically how to optimise movements over the junction, it was immediately obvious that it had worked out the principles programmers would need to write an ARS program and given them the various formulae they would have to incorporate.A year or so back I was following links around the web (as you do) and was reading about the BR Research setup, whose staff included mathematicians. There was a link to a paper read at one of the leading mathematical societies, the mathematics of regulating trains at a simple double junction. Seriously; that was worth reading at that level. So there's more to this signalling stuff than meets the eye.That could have been my old boss, Harry Ryland. Log in to reply |
| Getting the best out of ARS 09/09/2020 at 18:16 #131448 | |
|
GeoffM
6376 posts |
kbarber in post 131441 said:GeoffM in post 131418 said:I don't know where he started but he worked for BR Research by at least the 1970s, working on ARS and SSI - possibly IECC as well, but maybe not directly. Some point after that he joined Westinghouse Signals, which was where I met him (my boss's boss). He then started TRE in 2000 with two other directors and I was the first employee that *cough* happened to hear about this new company... Anyway, clever bloke and certainly mathematically minded.kbarber in post 131326 said:I'm quite sure the paper I came across was 1970s or even late '60s. It was totally mathematics, apart from explaining the layout of a double junction and the possible moves and conflicts. It had been read, as I say, at a mathematical society (maybe even The Mathematical Society of the UK - though I don't know if that's its proper name), which I imagine is pretty much equivalent to being published in a peer reviewed journal. And although it (as far as I can remember) contented itself with working out mathematically how to optimise movements over the junction, it was immediately obvious that it had worked out the principles programmers would need to write an ARS program and given them the various formulae they would have to incorporate.A year or so back I was following links around the web (as you do) and was reading about the BR Research setup, whose staff included mathematicians. There was a link to a paper read at one of the leading mathematical societies, the mathematics of regulating trains at a simple double junction. Seriously; that was worth reading at that level. So there's more to this signalling stuff than meets the eye.That could have been my old boss, Harry Ryland. I've just found our regulation strategies document. Obviously I can't share it, but suffice to say there are 30 pages of psuedo code and maths - and that's just the regulation part of ARS. The reference made earlier to metro regulation is likely very different - certainly I've read other maths-based papers on metro regulation and they're very heavy on the theory; very light on the practicalities of implementing it on a real metro with real trains and real passengers. SimSig Boss Log in to reply |
| Getting the best out of ARS 09/09/2020 at 22:36 #131452 | |
|
Jan
906 posts |
Maybe this one? (If not, it certainly sounds very similar to what you're remembering)
Two million people attempt to use Birmingham's magnificent rail network every year, with just over a million of them managing to get further than Smethwick. Log in to reply The following user said thank you: ozrail |
| Getting the best out of ARS 10/09/2020 at 08:26 #131468 | |
|
kbarber
1743 posts |
Jan in post 131452 said:Maybe this one? (If not, it certainly sounds very similar to what you're remembering)That looks rather like it. Likewise the glazed eyes when I try to read it. And the author's affiliation is RTC as well. Thank you Jan. Log in to reply |