Re: [talk-au] Incorrect entry to BP service station
On Wed, Jan 13, 2010 at 1:54 PM, Roy Wallace wrote: > But it's relevant to anyone who is placing POI's. Where should they be > placed - floating inside the building or on the building outline (at > the entrance)? Floating inside. Placing it on the building outline seems wrong because it's no longer centered, and how would you choose which was the "main entrance" anyway? Actually the POI doesn't even have to be in a building, because the POI describes the whole piece of land, of which a building might be a small (or non-existent) part. So, I don't think putting POIs on boundaries would get you very far. Steve ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
2010/1/13 Roy Wallace : > On Wed, Jan 13, 2010 at 11:59 AM, John Smith > wrote: >> >>> On a related issue re: 2), in the case of multiple businesses >>> sharing a building (e.g. a typical AU shopping centre with entrances >>> on the outside of the building), is it advisable to place a POI node >> >> This is micro mapping > > But it's relevant to anyone who is placing POI's. Where should they be I'm not saying it isn't but it's getting well beyond the original intent of this thread. > placed - floating inside the building or on the building outline (at > the entrance)? > Ideally it would be nice to get the layouts for shopping centres and map them as if they were stand alone buildings, that way we could print out or use our own directory instead of trying to find one in the shopping centres. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
On Wed, Jan 13, 2010 at 11:59 AM, John Smith wrote: > >> On a related issue re: 2), in the case of multiple businesses >> sharing a building (e.g. a typical AU shopping centre with entrances >> on the outside of the building), is it advisable to place a POI node > > This is micro mapping But it's relevant to anyone who is placing POI's. Where should they be placed - floating inside the building or on the building outline (at the entrance)? ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
2010/1/13 Roy Wallace : > Good question. I'm not sure. OSM's data should at least include 1) > service roads (to decrease the distance between the POI and the > nearest road, which should decrease ambiguity), and 2) entrances (to > resolve ambiguity in the case of more than one very close roads). The point of the highway=service is to map what's on the ground, not just to decrease the distance, as I said before some of these are one way so this is valuable information. > On a related issue re: 2), in the case of multiple businesses > sharing a building (e.g. a typical AU shopping centre with entrances > on the outside of the building), is it advisable to place a POI node This is micro mapping, some people have already come up with some interesting way to map indoors, I just wish I could get my head round the math needed to utilise an accelerometer and GPS in combination for when GPS degrades... ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
On Wed, Jan 13, 2010 at 11:27 AM, John Smith wrote: > > If we're mapping what's on the ground most POIs won't physically be > connected to the road way, servos are an obvious exception. > > If there is a problem with other POIs, where is the problem? > > * OSM's data > * the pre-processor > * the device/software that does routing Good question. I'm not sure. OSM's data should at least include 1) service roads (to decrease the distance between the POI and the nearest road, which should decrease ambiguity), and 2) entrances (to resolve ambiguity in the case of more than one very close roads). On a related issue re: 2), in the case of multiple businesses sharing a building (e.g. a typical AU shopping centre with entrances on the outside of the building), is it advisable to place a POI node *on* the building outline, rather than floating in the middle of the building? This seems sensible to me, as this is equivalent to tagging the location of the sign with the shop's name and/or the location of the entrance to the shop. Another advantage is that if the building is connected to the road grid, or a footway, etc., this may make it easier to route directly to the POI. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
2010/1/13 Roy Wallace : > On Wed, Jan 13, 2010 at 9:04 AM, Sam Couter wrote: >> >> Trying to drive or route to disconnected nodes is >> nonsensical. > > A question, then: what proportion of OSM POI's are disconnected? > Should we be taking steps (in terms of mapping guidelines) to ensure > POI nodes and buildings (and anything else likely to be a routing > target) are connected to the road grid? If we're mapping what's on the ground most POIs won't physically be connected to the road way, servos are an obvious exception. If there is a problem with other POIs, where is the problem? * OSM's data * the pre-processor * the device/software that does routing ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
On Wed, Jan 13, 2010 at 9:04 AM, Sam Couter wrote: > > Trying to drive or route to disconnected nodes is > nonsensical. A question, then: what proportion of OSM POI's are disconnected? Should we be taking steps (in terms of mapping guidelines) to ensure POI nodes and buildings (and anything else likely to be a routing target) are connected to the road grid? Presumably this has come up before - but this question seems to be the key issue here. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
2010/1/13 Sam Couter : > Left turn from Falcon Street onto Pacifiy Highway. If you switch to sat view you can see there is a seperate lane for that which makes the turn less than 90 degrees. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
2010/1/13 Roy Wallace : > On Tue, Jan 12, 2010 at 9:17 PM, John Smith wrote: >> >> In this case you have to assume the turn restictions are completely >> ignored, which if that is the case is there really any point in >> putting them in in the first place to tell stupid things how to do >> their jobs better when the programming should be improved? > > "Improved" how? > Code is generated based on assumptions, so the assumptions need to be improved and the code updated for the new assumptions. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
Roy Wallace wrote: > So the routing software has to approximate the target node with some > other node that *is* connected... and if the router does this > approximation sub-optimally, this is a problem with the router, right? How can it possibly know? This is garbage in, garbage out. When asked an impossible question, the answer is gibberish. The router has no way of knowing that the node is connected to two nearby ways rather than just the nearest, which is the approximation I'd expect. Service roads seem like the perfect solution: They're unabiguous and exactly correct. Trying to drive or route to disconnected nodes is nonsensical. -- Sam Couter | mailto:s...@couter.id.au OpenPGP fingerprint: A46B 9BB5 3148 7BEA 1F05 5BD5 8530 03AE DE89 C75C signature.asc Description: Digital signature ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
On Tue, Jan 12, 2010 at 9:17 PM, John Smith wrote: > > In this case you have to assume the turn restictions are completely > ignored, which if that is the case is there really any point in > putting them in in the first place to tell stupid things how to do > their jobs better when the programming should be improved? "Improved" how? ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
On Tue, Jan 12, 2010 at 8:46 PM, Richard Colless wrote: > > Has anyone else encountered this situation? Isn't this a problem with the router? I.e. if the target node is *unconnected* to the road network, then routing along the road network in order to arrive at the target node is technically *impossible*, right? So the routing software has to approximate the target node with some other node that *is* connected... and if the router does this approximation sub-optimally, this is a problem with the router, right? ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
John Smith wrote: > KISS... > > In this case you have to assume the turn restictions are completely > ignored, which if that is the case is there really any point in > putting them in in the first place to tell stupid things how to do > their jobs better when the programming should be improved? I disagree with your assertion that turns through acute angles are impossible, or even rare. One example: http://maps.google.com/maps?f=q&source=s_q&hl=en&geocode=&q=crows+nest,+nsw&sll=37.0625,-95.677068&sspn=40.681389,79.013672&ie=UTF8&hq=&hnear=Crows+Nest+New+South+Wales,+Australia&ll=-33.82794,151.2008&spn=0.005036,0.009645&z=17 Left turn from Falcon Street onto Pacifiy Highway. -- Sam Couter | mailto:s...@couter.id.au OpenPGP fingerprint: A46B 9BB5 3148 7BEA 1F05 5BD5 8530 03AE DE89 C75C signature.asc Description: Digital signature ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
2010/1/12 Roy Rankin : > My experience of many years with both people and computers is that > people often show that common sense is not that common and that > computers have none at all. It may be just a fringe case, but I believe > that if a turn is not prohibited then sooner or later a route will be > drawn the uses the turn even if it does not make sense. Yes well that's a diff matter... > I do not believe this attitude will result in the best routing > experience possible. KISS... In this case you have to assume the turn restictions are completely ignored, which if that is the case is there really any point in putting them in in the first place to tell stupid things how to do their jobs better when the programming should be improved? ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
On 01/12/2010 07:08 PM, Ross Scanlon wrote: > Likewise don't put roads that are on the same pavement on different layers > it's not mapping what's on the ground. OK, I have re-read the layers tag on the map features page and it says "This is intended to indicate actual physical separation, shouldn't be used just to influence rendering order." I thus conclude that using layers is the wrong approach. We now have 4 possible ways to map and in all cases crossing ways have a common node. 1> use no restrictions and assume oblique angle turns will not be made. 2> use highly schematic representations. 3> break ways at every node with appropriate turn restriction. 4> turning lane is an only_straight_on via in the relation. On 01/12/2010 07:07 PM, John Smith wrote: > 2010/1/12 Roy Rankin: >> The example I gave was a simplification to demonstrate the issues. From >> Johns response I did realize I had gotten the directions which I assumed >> is why it did not make sense to him. If you want to visualize the >> situation look at the link I supplied >> (http://www.openstreetmap.org/?lat=-33.82907&lon=151.21494&zoom=17&layers=0B00FTF >> ) and consider east bound traffic on Falcon Street turning right to go > > Only a couple of common sense things, cars normally don't make turns> > 90 degrees, so you generally don't need turning restirctions in this > cases. My experience of many years with both people and computers is that people often show that common sense is not that common and that computers have none at all. It may be just a fringe case, but I believe that if a turn is not prohibited then sooner or later a route will be drawn the uses the turn even if it does not make sense. > > Secondly, I don't think it looks legal for cars to go straight if they > are coming off the motorway as is the case with way IDs #47063022 and > #47063028, the arrows indicate turning only. I did not map this and my first reaction was similar to yours. When I looked more closely the arrows have buses excepted and the straight on way is tagged as psv=yes. There is also a dashed line to guide the buses to rejoin the motorway. > > Simply setting most of those turns to be one way will be sufficient > for routing (and humans for that matter) to figure it out. I do not believe this attitude will result in the best routing experience possible. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
John Smith wrote: 2010/1/12 Richard Colless : I thought of that. I just don't like using sledgehammers on walnuts. The service road exists in reality, so it's just a matter of improving the map to reflect reality more closely. In fact a lot of servos have one ways through them, so this additional information can be very useful. I'll put in a couple of service roads and try it out next time I have a delivery pickup in the area. Has anyone else encountered this situation? ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
On Tue, Jan 12, 2010 at 3:57 PM, wrote: > On Mon, 2010-01-11 at 21:44 +1100, Roy Rankin wrote: > With the resolution and newness of Nearmap images Freeway/Motorway >> interchanges are being mapped in detail rather than schematically. This >> brings up the issue on how to best do this. >> >> ... >> >> (http://www.openstreetmap.org/?lat=-33.82907&lon=151.21494&zoom=17&layers=0B00FTF >> ) The mapper did not realise that ways with turn restrictions must break >> at the via node. Unfortunately the current tested JOSM does not handle >> turn restriction relations properly when a way is split and much effort >> is required to fix up the relations after a way split. As each way had >> quite a few turn restrictions I realized that after splitting the ways >> it would require some time to sort out the relations. I thus used method >> 2 above and deleted all the relations. >> >> Regards, >> Roy Rankin >> > > Good point there Roy, > > This intersection was my doing, I wasn't 100% sure how to do turn > restriction relations that well, and OSM's documentation for straight_on > isn't clear. > http://wiki.openstreetmap.org/wiki/Relation:restriction > > That is a particularly messy intersection to deal with: > http://www.openstreetmap.org/?lat=-33.829754&lon=151.214575&zoom=18&layers=B000FTF > > Personally I'd prefer to use turn restriction relations rather than just > using different layer tags for these complex intersections, as it seems > to be more the "proper way" of doing it. > > After having a look at your handywork fixing up that intersection Roy, > it does indeed make reading the intersection much easier in JOSM. > > But indeed as you say, just using layers, one way roads, and some > non-connecting ways is in itself a rather elegant way of doing things. > > Would anyone else like to comment? I'm very much in favour of the turn restrictions over the layering. It is physically possible to make all these weird turns, it's just not legal to do so. Therefore I would prefer to map it this. cheers > > Regards, > Rhubarb. > -- > > cam_...@fastmail.fm > > -- > http://www.fastmail.fm - Access your email from home and the web > > > ___ > Talk-au mailing list > Talk-au@openstreetmap.org > http://lists.openstreetmap.org/listinfo/talk-au > -- Franc ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
2010/1/12 Richard Colless : > I thought of that. I just don't like using sledgehammers on walnuts. The service road exists in reality, so it's just a matter of improving the map to reflect reality more closely. In fact a lot of servos have one ways through them, so this additional information can be very useful. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
John Smith wrote: 2010/1/12 Richard Colless : Any suggestions? It might need a service road connecting the BP node to the surround roads. I thought of that. I just don't like using sledgehammers on walnuts. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Incorrect entry to BP service station
2010/1/12 Richard Colless : > Any suggestions? It might need a service road connecting the BP node to the surround roads. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
[talk-au] Incorrect entry to BP service station
I've been testing routing instructions on my Garmin for routes that I use regularly - makes great quality control, as it shows up badly drawn roundabouts, incorrect streets, all sorts of rubbish. A while back I tried to navigate to a BP service station from only a few hundred metres away. You will need to open up this link to understand the problem: http://www.openstreetmap.org/?lat=-33.75907&lon=150.7867&zoom=17 Keep it open for reference while reading the next few paragraphs, or they won't make sense. I was located at the eastern end of Kurrajong Road, close to the Boronia Road junction. I asked for routing instructions to the BP service station at the intersection of Glossop and Kurrajong. The first time I tried this, the instructions were nonsense, and I found out when I got home that the BP had been placed in the middle of a median strip. I moved it to its correct location when I got home, and tried the test again next time I was in the area. The route instructions asked me to travel to west to Birch Street, make a right turn, proceed north to Debrincat Avenue, then turn left into Glossop and proceed south to the service station -travel right round the block instead of using the service station entrance in Kurrajong Road. There are actual entrances in both Kurrajong and Glossop, and I thought that placing the node near the corner would make the router aware of this, but obviously not. Any suggestions? Richard ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
2010/1/12 Ross Scanlon : > Likewise don't put roads that are on the same pavement on different layers > it's not mapping what's on the ground. > > I know it's difficult with the current routers but we need to add the data > correctly then get the routers fixed to correctly work with the turn > and/or lane restriction relations. Who ever mapped it seems to have gone over board and mapped it too much, on the other hand there is physical lanes as well, the junction I posted a link to looks very similar to the one in question and it is mapped correctly with the over pass mapped as a bridge etc, and the motorway mapped not as a tunnel etc. ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
> There is a third solution which I have seen. Typically the interchange > ways cross 2 other ways. Normally I think of the via in a turn relation > being a node. I am not sure the routers work this way, but if the right > turn lane is a separate way with a common nodes with ways it crosses > (but turns are not allowed).The right hand turn then becomes a separate > way as the on-ramp. This is the same issue we discussed a few weeks ago where lanes are depicted with separate ways. IMO this is incorrect and should not be used. Likewise don't put roads that are on the same pavement on different layers it's not mapping what's on the ground. I know it's difficult with the current routers but we need to add the data correctly then get the routers fixed to correctly work with the turn and/or lane restriction relations. -- Cheers Ross ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au
Re: [talk-au] Mapping interchanges levels or turn restrictions
2010/1/12 Roy Rankin : > The example I gave was a simplification to demonstrate the issues. From > Johns response I did realize I had gotten the directions which I assumed > is why it did not make sense to him. If you want to visualize the > situation look at the link I supplied > (http://www.openstreetmap.org/?lat=-33.82907&lon=151.21494&zoom=17&layers=0B00FTF > ) and consider east bound traffic on Falcon Street turning right to go Only a couple of common sense things, cars normally don't make turns > 90 degrees, so you generally don't need turning restirctions in this cases. Secondly, I don't think it looks legal for cars to go straight if they are coming off the motorway as is the case with way IDs #47063022 and #47063028, the arrows indicate turning only. Simply setting most of those turns to be one way will be sufficient for routing (and humans for that matter) to figure it out. I saw plenty of these types of over passes in the US, if you want inspiration look there. eg: http://www.openstreetmap.org/?lat=40.10907&lon=-83.09151&zoom=17&layers=B000FTF ___ Talk-au mailing list Talk-au@openstreetmap.org http://lists.openstreetmap.org/listinfo/talk-au