logo separator

[mkgmap-dev] type=through_route relations and mkgmap r3743

From News news at pointdee.co.uk on Tue Jan 10 15:24:32 GMT 2017

Hi Gerd. It's been 12 months since I moved away from here so I have no 
idea what the actual results are now but please see my comments inline 
with the assumption that you follow the road you are on as indicated by 
the road markings unless you are told to turn off which is how I was 
taught to drive and have always driven

On 01/10/2017 03:04 PM, Gerd Petermann wrote:
> Hi Paul,
>
> thanks for the details.
> I've created maps for the two locations with r3746. Note that the type=through_route have no effect on the map in this version.
> The results:
> ACD gives "Keep left onto  A56"

Personally I would prefer "Turn left". To me a "Keep left" says "stay in 
the left hand lane" whereas in this situation you turn left out of the 
left hand lane in order to stay on the A56. However something is  better 
than nothing which is what happened previously

> ACB gives "Keep right onto  A682"

Is acceptable but technically incorrect as you can stay in the left hand 
lane if you wish. There is no need for any instruction here

>
> The 2nd example gives "Keep left onto Blackburn Road (B6527)"

At this junction that is fine as the road does indeed fork

>
> I think these hints are completely okay. For the last example there is a  "Turn right onto Smith Lane" instruction.
> I'd prefer a "Keep right onto Smith Lane"  at this point, but the important part is that there is a routing hint.
>

I'd prefer nothing. The assumption is that you follow the road you are 
on as indicated by the road markings unless you are told to turn off

> So, I think it is okay when mkgmap continues to ignore these relations as the img format has no native support for them,
> the only thing mkgmap can do is to store "wrong" heading" values, e.g. for the last case it could change the initial heading of
> B-D from 90 degrees to 15 degrees so that  a "Keep right onto Smith Lane"  is more likely, but that would also mean a slightly
> wrong  time calculation (no penalty for the sharp turn).

Your results seem better than the last time I used a map in this 
location which didn't have the through_routes

Thanks

Paul

>
> Gerd
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von News <news at pointdee.co.uk>
> Gesendet: Dienstag, 10. Januar 2017 15:19:16
> An: Development list for mkgmap
> Betreff: Re: [mkgmap-dev] type=through_route relations and mkgmap r3743
>
> The wanted effect is to fix incorrect route hints when the road layout
> in OSM is correct but ambiguous. Personally I see a through_route as
> being no different to no_left_turn or no_right_turn or similar
> restrictions in that it gives a routing engine the full information it
> needs to be able to make correct decisions and issue correct
> instructions. If we didn't add a no_left_turn at a junction where you
> can't turn left how is a routing engine to know that? It's purpose
> therefore is to issue a turn warning when you need to turn off a highway
> or to not give a turn warning when one isn't needed. The proposal can
> still be found at
> http://wiki.openstreetmap.org/wiki/Proposed_features/through_route but
> I'll try to explain it here
>
> Consider this
>
> A     B
>   \    |
>    \   | A682
> A56\  |
>      \ |
>       \|C
>        |
>        |
>    A56 |
>        |
>        D
>
> This is a real world example and is the one I used in the proposal
> (http://www.openstreetmap.org/#map=18/53.68901/-2.30585). The lanes here
> are marked as the (hopefully) attached jpg. Therefore if you're heading
> north on the A56 and want to stay on the A56 you have to make a left
> turn. Without the through_route relation a routing engine probably can't
> determine that you need to make a left turn at this point as presumably
> the angle between this and the A682 is too small. Before I added a
> through_route at this point then if I was driving north on the A56 and
> wanted to head towards Accrington (staying on the A56 but needing to
> make a left turn) then I received no indication that I needed to turn
> left at this junction and therefore would end up in Rawtenstall.
>
> Conversely if you're heading north on the A56 and want to head towards
> Rawtenstall then without the through_route you are told to stay right to
> join the A682. This is incorrect as you do not need to keep right you
> just need to continue in the lane you are in but not as bad as missing a
> turn.
>
> In short....
> If you're travelling D->C->A then without a through_route you are NOT
> told to turn left when you should be even though that's the way the
> route is taking
> If you're travelling D->C->B then without a through_route you are told
> to stay right (or turn right, I can't remember which) when you need to
> just keep going along the carriageway you're on
>
> There is another example here
> (http://www.openstreetmap.org/#map=17/53.67505/-2.30756). If you're
> travelling north on the B6527 (Market Street) and want to stay on the
> B6257 (Blackburn Road) then you have to turn left otherwise you will end
> up on Burnley Road. The road markings here indicate that Market Street
> onto Burnley Road is the through_route and therefore if your route keeps
> you on the B6527 you need to turn left at this junction
>
> All the above are corrected with a through_route
>
> The second use for this is below but to me this is of lesser importance
>
>   A
>   |
>   |
>   |
>   |
> B------------D
>   |
>   |
>   |
>   |
>   C
>
> C->B->D is Smith Road for example. B->A is Jones Road.
> The road markings indicate that at point B Smith Lane makes a 90 degree
> turn so anyone driving along here (C->B->D) would automatically follow
> the road. However in this case you may be told to turn right onto Smith
> Road. Adding a through_route for C->B->D tells a routing engine that it
> only needs to issue a turn warning if you are turning onto Jones Road
>
> I hope that helps but if not then please ask
>
> Thanks
>
> Paul
>
> On 01/10/2017 11:22 AM, Gerd Petermann wrote:
>> Hi Paul,
>>
>> sorry , I still have no idea what the wanted effect is. If you think that my example is tagged correctly,
>> in what case should it change the routing hints and to what?
>> Or maybe point me to another example.
>>
>> thanks,
>> Gerd
>> ________________________________________
>> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von News <news at pointdee.co.uk>
>> Gesendet: Dienstag, 10. Januar 2017 11:55:06
>> An: mkgmap-dev at lists.mkgmap.org.uk
>> Betreff: Re: [mkgmap-dev] type=through_route relations and mkgmap r3743
>>
>> If I remember correctly the idea of a through_route was first raised by
>> Mark Burton a few years ago. It was intended to solve the problem of
>> TOTSO's
>> (https://en.wikipedia.org/wiki/Glossary_of_road_transport_terms#TOTSO)
>> which is a situation where you have to turn off the carriageway you are
>> currently travelling on in order to stay on the same named road. In
>> these circumstances it is (almost) impossible for any routing engine to
>> deduce that it needs to give you a routing hint in order for you to stay
>> on the road that you are currently on. At the time I had a couple of
>> these near me where I knew that my Garmin was effectively missing a turn
>> and hence you would end up travelling in the wrong direction. Adding a
>> through_route at these places fixes this problem.
>>
>> There is also a secondary use where a road e.g. Smith Road, bends 90
>> degrees and you are sometimes told to turn left/right to stay on Smith
>> Road even though there is no turning
>>
>> A few years ago I tried to get this accepted into the list of official
>> tags however the vote resulted in a draw and I didn't have chance to
>> pursue it further.
>>
>> I have now moved and there are none currently near me that I could use
>> to check if this is still working
>>
>> See also......
>> http://www.mkgmap.org.uk/pipermail/mkgmap-dev/2013q2/017429.html
>> http://www.mkgmap.org.uk/pipermail/mkgmap-dev/2010q1/007028.html
>>
>> Thanks
>>
>> Paul
>>
>> On 01/10/2017 08:31 AM, Gerd Petermann wrote:
>>> I found one that might be interesting:
>>> https://www.openstreetmap.org/relation/406379
>>>
>>> If I got that right the intention was to tell mkgmap that
>>> it should change the bearing (heading) values in the map so that the Garmin
>>> algo doesn't produce a
>>> hint like "Turn left onto Calle Botánico Rivas Mateos" when travelling from
>>> way 24647941 (Calle José Luis Cotallo) to way 24311294 (Calle Botánico Rivas
>>> Mateos).
>>> (this is what happens now)
>>>
>>> What would you want instead?
>>> @Carlos: You have created that relation in 2010, maybe you remember?
>>>
>>> Gerd
>>>
>>>
>>>
>>>
>>> Gerd Petermann wrote
>>>> Hi all,
>>>>
>>>> I've noticed that mkgmap still checks these relations for correctness but
>>>> they have no influence on the created map.
>>>> Part of the code that handled them was removed with r3649:
>>>> http://www.mkgmap.org.uk/websvn/revision.php?repname=mkgmap&rev=3649
>>>>
>>>> If your map was created with r3743 or later, please check if you find a
>>>> crossing with such a relation which where routing
>>>> hints are wrong. In that case I would try to add code for the support,
>>>> else some unused code can be removed.
>>>>
>>>> Gerd
>>>> _______________________________________________
>>>> mkgmap-dev mailing list
>>>
>>>> mkgmap-dev at .org
>>>
>>>> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>>>
>>>
>>>
>>>
>>>
>>> --
>>> View this message in context: http://gis.19327.n8.nabble.com/type-through-route-relations-and-mkgmap-r3743-tp5888958p5889077.html
>>> Sent from the Mkgmap Development mailing list archive at Nabble.com.
>>> _______________________________________________
>>> mkgmap-dev mailing list
>>> mkgmap-dev at lists.mkgmap.org.uk
>>> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>>>
>>
>> _______________________________________________
>> mkgmap-dev mailing list
>> mkgmap-dev at lists.mkgmap.org.uk
>> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>> _______________________________________________
>> mkgmap-dev mailing list
>> mkgmap-dev at lists.mkgmap.org.uk
>> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>>
>
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>



More information about the mkgmap-dev mailing list