[Routing] comparison of navigation-apps

[marcus.wolschon at googlemail.com]

On Fri, 20 Mar 2009 15:14:43 +0200, Nic Roets <nroets at gmail.com> wrote:
> On Fri, Mar 20, 2009 at 2:34 PM, <marcus.wolschon at googlemail.com> wrote:
>> but couldn't figure it out. What algorithm are you using anyway?
Dijkstra
>> and A* cannot support restrictions on nodes and I could never find a
>>
> 
> "Dijkstra nodes" does not have to be "OSM nodes". They can be
state-machine
> states like "I'm at node 123, driving West on Main Road".

Interesting idea, that could get me on track again.
Sounds like just what I'm trying with my turn-restriction osmosis-task.
It would convert the intersection-node into a small graph of
oneway-streets.

However this turned out to have the problem of allowing a u-turn.
So:
* I cannot go left, instead I
* go right to the first node of the right street
* from that point I can reach the left street, so it does a u-turn.
that's why I halted further development on my version.

Can you give some more details about how you did it?
I found some code in libgosm.cpp but it's hard to understand without
more comments. Can you handle "no_u_turn"?

> Gosmore can search for any tag value, nearest matches first. So if I want
> to
> search for "1 Main Road, Lynnwood Park", I start by searching for
Lynnwood
> Park, then I search for Main Road. Then search for 1. I know it's not
> perfect, but it's not useless either.

Ah, that explains why I could not find it. Thank you.
Shall we label it "limited" then?

>> Well, we had 10(+1) offboard navigation-application in a plain list.
>> I merely expanded on that so the users have a chance to see what these
>> programs can do or at least what platform they are written for without
>> visiting and learning about each of them.
>> A major turndown for a user who simply looks for a navigation or
routing-
>> application to USE.
> 
> I agree. But it will be much better if someone actually downloads the
> applications and see if they work on a randomly chosen location in the
map,
> not just in the author's neighbourhood. For example not take to long to
> compute a route.

Good idea. I was to try out how far navit has advanced this weekend after
documenting the improved plugin-api in Traveling Salesman 0.9.8.
I figured as we have a routing mailing-list it was best to start with
posting
it here so the authors of these programs themself can help with the
collection
and correction if one gets a wrong impression when looking at the website,
code
and later (if possible) program. After all, they have to know it best. ;)

Marcus