[mkgmap-dev] Patch for better Autorouting

Tue Dec 1 11:42:47 GMT 2009

On 01.12.2009 12:38, Clinton Gladstone wrote:
> 2009/12/1 Felix Hartmann<extremecarver at googlemail.com>:
>    
>> As announced yesterday - here is a patch against the default-style-file to
>> heavily improve Autorouting for cars/motorcycles.
>>      
> So this patch, if I understand it correctly, creates duplicate ways at
> a higher resolution, but with a lower speed. Do you have any idea why
> this improves routing?
>    
Yes because of the fuc**d up Turn time Restrictions AND Garmin 
GPS/Mapsource trying far too hard to avoid sharp turns at higher speeds 
(which is a good idea, but not well implemented). The resolution does 
not matter - I put it at 24 because there is no need for them to be 
visible, and routing only depends on resolution=24.

This is a dirty hack, but the easiest to do, see below alternative 
solutions (wrote that in another thread here):

"Actually I have found a method to increase routing distance in 
Mapsource by 30-100% -due to the functionality introduced yesterday, 
however it blows up the map size by at least 10%. Will test it a bit 
more, and then add a patch for mkgmap default style. Based on my 
findings, I think we would really have to use the dp filter already 
BEFORE doing anything to get smoother/straighter roads AND work on 
junctions. I'ld guesstimate the following increases in routing distance 
(each thing on it`s own):

a) optimal - as discussed
Junction Angle and creating additional (invisble via typfile) "turn 
roads" ~ 500%
- Mapsize increase <5%

b) "Modest" patch:
Make a last 10m from each junction into any direction duplicate - 
decrease road_speed patch. >200%
- Mapsize increase <5%, could even be used together with a) but will 
probably show no further improvements then.

c)"dirty style-file patch"
Add slower duplicate roads for primary/trunks/motorways/motorway_links. 
~30-100%.
- Drawback mapsize increase ~10%. The more duplicates, the better, but 
the mapsize increase will be even bigger.
If either a) or b) is implemented this dirty hack will not be needed.

However even in comparison with c) - all previous attempts to increase 
routing quality seem to be pretty basic and unnoticable.2
> Cheers.
>    