Re: [OSM-talk] Removing redundant routing instructions
On 28/04/15 05:10, Bryce Nesbitt wrote: I'd call this mostly a routing presentation issue. If the road name is the same, I'd want any super sharp curve to warn me: Tight left in 100 meters, or 15mph left turn ahead. The very fact of the OSM geometry ought to be enough to calculate the necessary warning. Exactly my point, although that still leaves a few edge cases that would benefit as others have said. The discussion should be happening on the various router support lists and address their particular variations, but I think it is appropriate here to have the general guide lines that allow all routers to be able to rely on the one set of rules? -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On Tue Apr 28 10:10:00 2015 GMT+0100, Lester Caine wrote: On 28/04/15 05:10, Bryce Nesbitt wrote: I'd call this mostly a routing presentation issue. If the road name is the same, I'd want any super sharp curve to warn me: Tight left in 100 meters, or 15mph left turn ahead. The very fact of the OSM geometry ought to be enough to calculate the necessary warning. Exactly my point, although that still leaves a few edge cases that would benefit as others have said. The discussion should be happening on the various router support lists and address their particular variations, but I think it is appropriate here to have the general guide lines that allow all routers to be able to rely on the one set of rules? +1 Looking at some cases locally, there are certainly difference between the routers. Mapquest does seem to be better here. OSRM in particular does seem to reduce carefully drawn road geometry to a series of straight lines, which reduces its chances of getting turn instructions right in these cases. Phil (trigpoint ) -- Sent from my Jolla ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
Agree with that! On 2015-04-28 11:10, Lester Caine wrote: On 28/04/15 05:10, Bryce Nesbitt wrote: I'd call this mostly a routing presentation issue. If the road name is the same, I'd want any super sharp curve to warn me: Tight left in 100 meters, or 15mph left turn ahead. The very fact of the OSM geometry ought to be enough to calculate the necessary warning. Exactly my point, although that still leaves a few edge cases that would benefit as others have said. The discussion should be happening on the various router support lists and address their particular variations, but I think it is appropriate here to have the general guide lines that allow all routers to be able to rely on the one set of rules? ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
FWIW, I just wanted to fix the situation to give you an example, and someone else was faster, it is already fixed in the way I did suggest above: http://www.openstreetmap.org/#map=19/52.45290/-1.48908 Cheers, Martin ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On Sun Apr 26 12:35:57 2015 GMT+0100, Rob Nickerson wrote: Hi all, In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. Although the road continues round the bend SatNav systems often think it is a junction and tell you to turn right/left in 100 yards/meters. I wonder whether it is possible to indicate this in OpenStreetMap so that routing engines can omit this redundant instruction. == Example picture == https://drive.google.com/file/d/0B6J5ZA1hu93bZmx2NTIxaHdfMUE/view?usp=sharing In the example Oban Road [1] turns to the right to become the northern section of Sydnall Road. All main routers tell you to turn right. In my opinion this is a redundant instruction (or could be better worded). I've tried to add extra nodes so that the road naturally bends but the main routing engines still tell you to turn. == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? And more importantly, if you need to turn off onto the minor road going straight ahead it remains 'silent'. I have occasionally used a through_route relation in these cases, but lack of support from routers does make it seem futile. Much like the via way relation, that one is so needed too. Phil (trigpoint) -- Sent from my Jolla ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On 27/04/15 10:45, p...@trigpoint.me.uk wrote: == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? And more importantly, if you need to turn off onto the minor road going straight ahead it remains 'silent'. I have occasionally used a through_route relation in these cases, but lack of support from routers does make it seem futile. Much like the via way relation, that one is so needed too. Why do I find that confusing? Currently in general the directions ignore corners even if there are roads going off those corners. The complaint is about 'extra' directions where the corner is actually the main road and the straight on branch is the turning. If the directions remain silent one would in some conditions get confused so the safe thing to do is announce both? The correct wording of the the directions should be perhaps 'road bares right' and 'go straight on' rather than 'turn right' but that does need the perhaps missing through_route information? As I said, the inclusion of road names in OSMAND while useful at times does get similarly annoying when a 'continue on Axxx' would suffice. In this case the road id provides the through_route information ... one remains on the same road ... and the straight on road has a different one which may just be 'minor road'. -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
2015-04-27 10:52 GMT+02:00 Cartinus carti...@xs4all.nl: If you reread the original mail, then you'll see he already tried this himself and it did not work. Maybe this is caused by geometry simplication in the router? If you have a detailed look at the overlay you can see that the routing geometry is not following completely the rendered geometry: http://www.openstreetmap.org/directions?engine=osrm_carroute=52.45362%2C-1.48598%3B52.45341%2C-1.48944#map=18/52.45311/-1.48788layers=Q On the other hand, OSRM test website (with apparently different geometry) also show a turn slight right at this spot: http://map.project-osrm.org/?hl=enloc=52.453163,-1.488334loc=52.453129,-1.489063z=18center=52.453214,-1.489010alt=0df=0re=0ly=-1171809665 I guess in this case I wouldn't mind if I got the turn slight right instruction, in the end that's what you have to do, but I agree it would be better to get stay on the road and turn slight right or something like this. The other case (main road turns right but you have to keep straight) is more important to be fixed. Cheers, Martin ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
2015-04-26 13:35 GMT+02:00 Rob Nickerson rob.j.nicker...@gmail.com: I wonder whether it is possible to indicate this in OpenStreetMap so that routing engines can omit this redundant instruction. == Example picture == https://drive.google.com/file/d/0B6J5ZA1hu93bZmx2NTIxaHdfMUE/view?usp=sharing I think you might be able to fix this with micromapping. If you add the detailed shape (i.e. the curve like it is signed on the road) of the junction topology it might be solved. It is a general issue in many cases that the junctions don't get modelled accurately (you can't determine from the drawn shape which road is continuous and which gets discharged into the other, because of oversimplification / bad generalization). Cheers, Martin ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
If you reread the original mail, then you'll see he already tried this himself and it did not work. On 27-04-15 10:46, Martin Koppenhoefer wrote: FWIW, I just wanted to fix the situation to give you an example, and someone else was faster, it is already fixed in the way I did suggest above: http://www.openstreetmap.org/#map=19/52.45290/-1.48908 Cheers, Martin ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk -- --- m.v.g., Cartinus ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On 26/04/2015 12:35, Rob Nickerson wrote: In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. One simple way of representing this situation is to place give_way nodes on the subsidiary roads. Whether any routers or renderers make use of these to deduce that a particular route through the junction is the 'through route', I don't know. -- Steve --- This email has been checked for viruses by Avast antivirus software. http://www.avast.com ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On 27/04/15 09:06, Steve Doerr wrote: In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. One simple way of representing this situation is to place give_way nodes on the subsidiary roads. Whether any routers or renderers make use of these to deduce that a particular route through the junction is the 'through route', I don't know. It does irritate me that OSMAND announces 'turn right onto A46' when I'm currently driving down the A46, and 'turn slightly left onto M5' when approaching a slip road to 'turn right onto A46'. There is NOTHING wrong with the tagging in OSM, it is purely how OSMAND handles each case which needs further work. Other satnav's do the same sort of thing in different areas, and around here many roads don't have road numbers, so when moving between sections of what is essentially the same road one gets directions about the junction. All that was missing from the tagging was an indication that it is the same road, so yes signs on the side junctions might help but it's up to the third party software to interpret that. The opposite side of the coin is when 'tomtom' tells me it's 50 miles to the next junction and I know I've got two or three major interchanges where I need to merge with other traffic. It just depends what one considers to be redundant. The example given says 'the white lines mark it as such', but if those white lines are warn or obscured by mud? The extra direction can ACTUALLY be helpful if it's a route one has not been before. It only becomes irritating when one is hearing it for the n'th time ;) -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
The trouble with nodes is that they are non-directional. Junctions in quick succession, and lane-dependent give-ways could make a challenging scenario for a program to try and make sense of. Why not tag it explicitly instead of leaving it to heuristics which (by definition) will not always get it right? //colin On 2015-04-27 10:06, Steve Doerr wrote: On 26/04/2015 12:35, Rob Nickerson wrote: In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. One simple way of representing this situation is to place give_way nodes on the subsidiary roads. Whether any routers or renderers make use of these to deduce that a particular route through the junction is the 'through route', I don't know. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
As I understand it, there is an implied direction in that the convention is that the give_way node applies to the nearest intersection involving the way. But yes, I can see that involves extra computation. Steve On 27/04/2015 09:51, Colin Smale wrote: The trouble with nodes is that they are non-directional. Junctions in quick succession, and lane-dependent give-ways could make a challenging scenario for a program to try and make sense of. Why not tag it explicitly instead of leaving it to heuristics which (by definition) will not always get it right? //colin On 2015-04-27 10:06, Steve Doerr wrote: On 26/04/2015 12:35, Rob Nickerson wrote: In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. One simple way of representing this situation is to place give_way nodes on the subsidiary roads. Whether any routers or renderers make use of these to deduce that a particular route through the junction is the 'through route', I don't know. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk --- This email has been checked for viruses by Avast antivirus software. http://www.avast.com ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On Mon Apr 27 15:07:45 2015 GMT+0100, Marc Gemis wrote: As long as the name (or the ref/int_ref) of the street remains the same, I think the router should be able to give other messages than turn right. There is no need for an additional relation IMHO. There is often no ref, or name. If there is a name it will often change. Phil (trigpoint ) On Mon, Apr 27, 2015 at 12:06 PM, Lester Caine les...@lsces.co.uk wrote: On 27/04/15 10:45, p...@trigpoint.me.uk wrote: == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? And more importantly, if you need to turn off onto the minor road going straight ahead it remains 'silent'. I have occasionally used a through_route relation in these cases, but lack of support from routers does make it seem futile. Much like the via way relation, that one is so needed too. Why do I find that confusing? Currently in general the directions ignore corners even if there are roads going off those corners. The complaint is about 'extra' directions where the corner is actually the main road and the straight on branch is the turning. If the directions remain silent one would in some conditions get confused so the safe thing to do is announce both? The correct wording of the the directions should be perhaps 'road bares right' and 'go straight on' rather than 'turn right' but that does need the perhaps missing through_route information? As I said, the inclusion of road names in OSMAND while useful at times does get similarly annoying when a 'continue on Axxx' would suffice. In this case the road id provides the through_route information ... one remains on the same road ... and the straight on road has a different one which may just be 'minor road'. -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk -- Sent from my Jolla ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
As long as the name (or the ref/int_ref) of the street remains the same, I think the router should be able to give other messages than turn right. There is no need for an additional relation IMHO. m On Mon, Apr 27, 2015 at 12:06 PM, Lester Caine les...@lsces.co.uk wrote: On 27/04/15 10:45, p...@trigpoint.me.uk wrote: == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? And more importantly, if you need to turn off onto the minor road going straight ahead it remains 'silent'. I have occasionally used a through_route relation in these cases, but lack of support from routers does make it seem futile. Much like the via way relation, that one is so needed too. Why do I find that confusing? Currently in general the directions ignore corners even if there are roads going off those corners. The complaint is about 'extra' directions where the corner is actually the main road and the straight on branch is the turning. If the directions remain silent one would in some conditions get confused so the safe thing to do is announce both? The correct wording of the the directions should be perhaps 'road bares right' and 'go straight on' rather than 'turn right' but that does need the perhaps missing through_route information? As I said, the inclusion of road names in OSMAND while useful at times does get similarly annoying when a 'continue on Axxx' would suffice. In this case the road id provides the through_route information ... one remains on the same road ... and the straight on road has a different one which may just be 'minor road'. -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
I should have written then there is no need (with then when there is a name or ref that stays the same) in the other cases you need a relation. On Mon, Apr 27, 2015 at 4:18 PM, p...@trigpoint.me.uk wrote: On Mon Apr 27 15:07:45 2015 GMT+0100, Marc Gemis wrote: As long as the name (or the ref/int_ref) of the street remains the same, I think the router should be able to give other messages than turn right. There is no need for an additional relation IMHO. There is often no ref, or name. If there is a name it will often change. Phil (trigpoint ) On Mon, Apr 27, 2015 at 12:06 PM, Lester Caine les...@lsces.co.uk wrote: On 27/04/15 10:45, p...@trigpoint.me.uk wrote: == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? And more importantly, if you need to turn off onto the minor road going straight ahead it remains 'silent'. I have occasionally used a through_route relation in these cases, but lack of support from routers does make it seem futile. Much like the via way relation, that one is so needed too. Why do I find that confusing? Currently in general the directions ignore corners even if there are roads going off those corners. The complaint is about 'extra' directions where the corner is actually the main road and the straight on branch is the turning. If the directions remain silent one would in some conditions get confused so the safe thing to do is announce both? The correct wording of the the directions should be perhaps 'road bares right' and 'go straight on' rather than 'turn right' but that does need the perhaps missing through_route information? As I said, the inclusion of road names in OSMAND while useful at times does get similarly annoying when a 'continue on Axxx' would suffice. In this case the road id provides the through_route information ... one remains on the same road ... and the straight on road has a different one which may just be 'minor road'. -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk -- Sent from my Jolla ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
Won't work in the UK as there are plenty of cases where you have to give way and make a proper turn in order to stay on the same road name and/or ref. The concept even has a name - TOTSO which means Turn Off To Stay On. http://en.wikipedia.org/wiki/User:Martinvl/TOTSO You cannot reliably infer from the geometry and name/ref which road (or lanes) has to give way. The only way to improve the navigation instructions is by giving hints. This is not a routing question - nobody here is discussing which road is the best way to MyTown - it's about how you present the router's choices to the user. //colin On 2015-04-27 16:07, Marc Gemis wrote: As long as the name (or the ref/int_ref) of the street remains the same, I think the router should be able to give other messages than turn right. There is no need for an additional relation IMHO. m On Mon, Apr 27, 2015 at 12:06 PM, Lester Caine les...@lsces.co.uk wrote: On 27/04/15 10:45, p...@trigpoint.me.uk wrote: == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? And more importantly, if you need to turn off onto the minor road going straight ahead it remains 'silent'. I have occasionally used a through_route relation in these cases, but lack of support from routers does make it seem futile. Much like the via way relation, that one is so needed too. Why do I find that confusing? Currently in general the directions ignore corners even if there are roads going off those corners. The complaint is about 'extra' directions where the corner is actually the main road and the straight on branch is the turning. If the directions remain silent one would in some conditions get confused so the safe thing to do is announce both? The correct wording of the the directions should be perhaps 'road bares right' and 'go straight on' rather than 'turn right' but that does need the perhaps missing through_route information? As I said, the inclusion of road names in OSMAND while useful at times does get similarly annoying when a 'continue on Axxx' would suffice. In this case the road id provides the through_route information ... one remains on the same road ... and the straight on road has a different one which may just be 'minor road'. -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact [1] L.S.Caine Electronic Services - http://lsces.co.uk [2] EnquirySolve - http://enquirysolve.com/ [3] Model Engineers Digital Workshop - http://medw.co.uk [4] Rainbow Digital Media - http://rainbowdigitalmedia.co.uk [5] ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk [6] ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk [6] Links: -- [1] http://lsces.co.uk/wiki/?page=contact [2] http://lsces.co.uk [3] http://enquirysolve.com/ [4] http://medw.co.uk [5] http://rainbowdigitalmedia.co.uk [6] https://lists.openstreetmap.org/listinfo/talk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
I'd call this mostly a routing presentation issue. If the road name is the same, I'd want any super sharp curve to warn me: Tight left in 100 meters, or 15mph left turn ahead. The very fact of the OSM geometry ought to be enough to calculate the necessary warning. ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On 27 April 2015 at 13:52, Lester Caine les...@lsces.co.uk wrote: On 27/04/15 13:17, pmailkeey . wrote: Is the 'through route' and 'the same road' the same thing ? and does it mean that the road number stays the same or that you do not cross the white paint ? One of the routes I follow regularly is an old Roman cross road which is straight, but crosses several more major roads with give way or stop at every junction. It is the one B class road, but crosses white paint at every junction. So there is not a general rule that can be applied. The place 'through route' would be helpful is where a road bends at a junction, and the main route route is not straight on. In the example given, the road name changed and if there is no road reference to override that then either you announce the turn, or you need something else to switch it off? I think the answer is that less ambiguous terminology is needed. 'Turn off' and 'stay on' are ambiguous. Also, instructions relying on something else aren't good, e.g. Follow 'A1'. -- Mike. @millomweb https://sites.google.com/site/millomweb/index/introduction - For all your info on Millom and South Copeland via *the area's premier website - * *currently unavailable due to ongoing harassment of me, my family, property pets* TCs https://sites.google.com/site/pmailkeey/e-mail ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On 27/04/15 16:49, pmailkeey . wrote: On 27 April 2015 at 13:52, Lester Caine les...@lsces.co.uk mailto:les...@lsces.co.uk wrote: On 27/04/15 13:17, pmailkeey . wrote: Is the 'through route' and 'the same road' the same thing ? and does it mean that the road number stays the same or that you do not cross the white paint ? One of the routes I follow regularly is an old Roman cross road which is straight, but crosses several more major roads with give way or stop at every junction. It is the one B class road, but crosses white paint at every junction. So there is not a general rule that can be applied. The place 'through route' would be helpful is where a road bends at a junction, and the main route route is not straight on. In the example given, the road name changed and if there is no road reference to override that then either you announce the turn, or you need something else to switch it off? I think the answer is that less ambiguous terminology is needed. 'Turn off' and 'stay on' are ambiguous. Also, instructions relying on something else aren't good, e.g. Follow 'A1'. As I said ... 'turn onto A46' when one is already on the A46 ... As long a one has an identifier that matches before and after a junction, then the speech set CAN change. Where you have a traffic route that has road names then it should perhaps be optional to use the road names and while the 'A' route gives 'stay on', the change of road name gives the 'over junction to yyy'. With unnamed roads then something else would be needed to identify the through_route, but if that is an extra tag, or just 'give way' on the other connections just needs some agreement. While we don't 'tag for the routers', a few common ground rules that ensure that how a junction is navigate can be consistently 'routed' would not hurt and is just sensible data to be included in the database. From a rendering point of view it should allow the 'white lines' to be added even if they don't physically exist. While a give way should have road markings, often these days it IS only a sign. -- Lester Caine - G8HFL - Contact - http://lsces.co.uk/wiki/?page=contact L.S.Caine Electronic Services - http://lsces.co.uk EnquirySolve - http://enquirysolve.com/ Model Engineers Digital Workshop - http://medw.co.uk Rainbow Digital Media - http://rainbowdigitalmedia.co.uk ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
On 26 April 2015 at 12:35, Rob Nickerson rob.j.nicker...@gmail.com wrote: Hi all, In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. Although the road continues round the bend SatNav systems often think it is a junction and tell you to turn right/left in 100 yards/meters. I wonder whether it is possible to indicate this in OpenStreetMap so that routing engines can omit this redundant instruction. == Example picture == https://drive.google.com/file/d/0B6J5ZA1hu93bZmx2NTIxaHdfMUE/view?usp=sharing In the example Oban Road [1] turns to the right to become the northern section of Sydnall Road. All main routers tell you to turn right. In my opinion this is a redundant instruction (or could be better worded). I've tried to add extra nodes so that the road naturally bends but the main routing engines still tell you to turn. == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? Regards, Rob [1] http://www.openstreetmap.org/directions?engine=osrm_carroute=52.45362%2C-1.48598%3B52.45341%2C-1.48944#map=18/52.45332/-1.48771layers=Q ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk I think the instruction is sometimes required and it is therefore better to have it in than not. I'm sure without it, drivers would miss the turn from Holborn Hill into Moor Road https://www.openstreetmap.org/directions?engine=osrm_carroute=54.2119%2C-3.2791%3B54.2112%2C-3.2735#map=18/54.21158/-3.27626 despite what's left of the white lines indicating this is how the main route goes. -- Mike. @millomweb https://sites.google.com/site/millomweb/index/introduction - For all your info on Millom and South Copeland via *the area's premier website - * *currently unavailable due to ongoing harassment of me, my family, property pets* TCs https://sites.google.com/site/pmailkeey/e-mail ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
Rob Nickerson rob.j.nicker...@gmail.com writes: I'm not sure I get your point about hint for router versus aid for navigation. I suspect this may stem from the don't tag for the renderer rule. If we look at the end use case the aim is to get a routing engine that provides an optimal route with user friendly route instructions. I can't believe this is an easy tag and as such I would expect the routing developers to be raising issues they cannot solve via code alone. This is one area where I would like my SatNav not to spew redundant instructions. [not all directed at you of course] I think the text of the rejected proposal could have been written better. I think declaring this a hint for the router is not really the right characterization and led to trouble. We're talking about encoding facts about the world that are useful for multiple purposes. Also, don't tag for the renderer is about not putting wrong tags in because some renderer will make it look how you want. Describing something accurately so that it can be rendered is totally ok, and the main thing we do. What's really going on is that on the ground, it is (often) clear to a driver what continuing on this road means, vs turning. This can be because of signs, or because of subtle geometry of curb cuts, or the widths of the continue vs turn roads, or various other clues. Somtimes, many of these clues can't be figured out from aerials, and certainly not From road vectors. So it makes sense to have a way to tag this so that the map data captures this on-the-ground truth. Often it's obvious from the geometry (and the obvious answer is right), and those cases don't need tags. I think for now we should avoid getting into rules about when it's not needed, and be ok with people adding the tags if they think it's confusing. I view this as similar to turn restrictions, except that instead of telling you what you can't do, it's documenting what it means to not turn. I would suggest something like turn_description=straight, or maybe =none, to be used on the from/to ways at any junction where the look-at-the-map-and-obvious-guess answer isn't right. They should be directional because perception is going to be different, and sometimes you'll need both. There's a further question about whether routers should announce. I'd say that if the road turns 90 degrees, it should, but it should say turn left to stay on Route 2 rather than nothing. I guess this is another wrinkle in tagging, and again is a fact about the world more than a router kludge. So I come down to turn_description=straight (road continues, obvious to driver) turn_description=stay_on_road (road continues, not obvious to driver) pgpeAG4FEkKW0.pgp Description: PGP signature ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
The difference between routing and navigation is that the routing algorithm will work out which road you need to be on, but it is the navigation aspect which makes translates the routing graph to useful instructions for a human. If the main road does a 90 degree left at a T-junction, something has to work out whether to say follow the road through the bend or turn left or something else or nothing at all. Anyone who implies that geometry alone is enough to make that decision, is confusing the two concepts IMHO. Arguments that it is not needed for routing are possibly correct, but short-sighted. The through_route idea doesn't change anything for the routing (although routing algorithms may add a time penalty for give-ways or sharp bends if they wish) but it adds a hint as to how best to describe the next move for the driver. A motorway exit is also an aid to navigation in that it affects only the (spoken) instructions, and not the route chosen by the routing engine. I would be in favour of reviving the old proposal. //colin On 2015-04-26 15:26, Rob Nickerson wrote: There already is a through_route relation, to show the path of the through route. It might not be well documented, but it is used (I believe)by mkgmap. There was a proposal, which was eventually rejected: http://wiki.openstreetmap.org/wiki/Proposed_features/through_route [2] IMHO it was rejected as it was seen as a hint for the router, not as an aid to navigation, and people don't understand the difference. Yeah I'm aware of that. In fact my example has been sat on my computer since that proposal and I've only just got back to looking at it!! This is in effect a revival of that proposal with a quite different example. I picked a different name as Through Route has a meaning in the UK - it means a route that takes you past a town whilst avoiding the congested city centre. If you think we should revive the through_route proposal then I'm happy with that instead. I'm not sure I get your point about hint for router versus aid for navigation. I suspect this may stem from the don't tag for the renderer rule. If we look at the end use case the aim is to get a routing engine that provides an optimal route with user friendly route instructions. I can't believe this is an easy tag and as such I would expect the routing developers to be raising issues they cannot solve via code alone. This is one area where I would like my SatNav not to spew redundant instructions. Best, Rob p.s. Is highway=motorway_junction a hint for router or an aid for navigation? ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk [1] Links: -- [1] https://lists.openstreetmap.org/listinfo/talk [2] http://wiki.openstreetmap.org/wiki/Proposed_features/through_route ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
There already is a through_route relation, to show the path of the through route. It might not be well documented, but it is used (I believe)by mkgmap. There was a proposal, which was eventually rejected: http://wiki.openstreetmap.org/wiki/Proposed_features/through_route IMHO it was rejected as it was seen as a hint for the router, not as an aid to navigation, and people don't understand the difference. On 2015-04-26 13:35, Rob Nickerson wrote: Hi all, In the UK (particularly in rural areas) it is common to find a road that turns 90 degrees to the left or right without a junction (that is the road just continues and white lines mark it as such). Meanwhile another road may come in from the other side with a 'give way' style junction. Although the road continues round the bend SatNav systems often think it is a junction and tell you to turn right/left in 100 yards/meters. I wonder whether it is possible to indicate this in OpenStreetMap so that routing engines can omit this redundant instruction. == Example picture == https://drive.google.com/file/d/0B6J5ZA1hu93bZmx2NTIxaHdfMUE/view?usp=sharing [2] In the example Oban Road [1] turns to the right to become the northern section of Sydnall Road. All main routers tell you to turn right. In my opinion this is a redundant instruction (or could be better worded). I've tried to add extra nodes so that the road naturally bends but the main routing engines still tell you to turn. == Question == Could we benefit from a new route relation? For example a route_continues relation? Would others find this useful? Regards, Rob [1] http://www.openstreetmap.org/directions?engine=osrm_carroute=52.45362%2C-1.48598%3B52.45341%2C-1.48944#map=18/52.45332/-1.48771layers=Q [3] ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk [1] Links: -- [1] https://lists.openstreetmap.org/listinfo/talk [2] https://drive.google.com/file/d/0B6J5ZA1hu93bZmx2NTIxaHdfMUE/view?usp=sharing [3] http://www.openstreetmap.org/directions?engine=osrm_caramp;route=52.45362%2C-1.48598%3B52.45341%2C-1.48944#map=18/52.45332/-1.48771amp;layers=Q___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] Removing redundant routing instructions
There already is a through_route relation, to show the path of the through route. It might not be well documented, but it is used (I believe)by mkgmap. There was a proposal, which was eventually rejected: http://wiki.openstreetmap.org/wiki/Proposed_features/through_route IMHO it was rejected as it was seen as a hint for the router, not as an aid to navigation, and people don't understand the difference. Yeah I'm aware of that. In fact my example has been sat on my computer since that proposal and I've only just got back to looking at it!! This is in effect a revival of that proposal with a quite different example. I picked a different name as Through Route has a meaning in the UK - it means a route that takes you past a town whilst avoiding the congested city centre. If you think we should revive the through_route proposal then I'm happy with that instead. I'm not sure I get your point about hint for router versus aid for navigation. I suspect this may stem from the don't tag for the renderer rule. If we look at the end use case the aim is to get a routing engine that provides an optimal route with user friendly route instructions. I can't believe this is an easy tag and as such I would expect the routing developers to be raising issues they cannot solve via code alone. This is one area where I would like my SatNav not to spew redundant instructions. Best, Rob p.s. Is highway=motorway_junction a hint for router or an aid for navigation? ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk