Detour points

[clegroux]

Hello,

I would like to solve a problem and don't know if it's possible to do it easily with PTV xServer.

Here is my problem

Let's take 2 locations A et B. It takes (for example) 30 minutes to go from A to B. I'm at point A and want to go to point B but I have 2 hours (for example) to do this trip.
How can I get all the points that I can reach within these 2 hours ? That's to say, how can I get all the points X that match this condition : (travel time between A and X) + (travel time between X and B) <= 2 hours ?

Charlie

[Joost]

What order of magnitude of possible candidate points are we talking about? A few? tenfold, hundredfold? etc

[clegroux]

Oh sorry I forgot to talk about one important thing

In fact I don't want to get the points that match the condition. I would like to get the area that contains all these points (like an isochrone).

[Bernd Welter]

You want something like the blue ellipse, the geometry?

"Distance" between A and B is smaller than the available horizon, so there is a search space to be reacheable

I assume we don't have a function for this but maybe you can use some simple (A:N) + (N:B) matrix requests?

[clegroux]

Yes exactly!

You're right I could use matrix requests but I would prefer to avoid calculating a distance matrix. Using a distance matrix means divide my map into a lot of squares (for me it's the best way to get a precise area). Do you have another simple way to solve my problem ?

You assume you don't have a function for calculating such a area. Do you plan to develop such a function ?

[Bernd Welter]

To be honest: the function you'd like to have is requested once every two years
Therefore I wouldn't expect this feature to come soon . . . Sorry for that.

Getting back to the functional requirement:
What do you mean by "slicing into rectangles"?

Are you doubting in performance or in quality of results?
I'm sure e can find at least some approaches of how to solve the given operational tasks...

Best regards
Bernd

Two 30min isochrones, checking for 1 hour.
Any point outside both isochrones will not be a candidate for the given task.
Any point inside both isochrones is reacheable within 60 minutes.
Attention: doesn't guarantee anything for points inside exactly one isochrone.

[clegroux]

Arf... So let's forget the function

First I'm doubting in quality of results. Yes, in your example, any point inside both isochrones is reachable within 60 minutes but we miss a lot of points (as you said, points inside exactly one isochrone may match the condition).

The only way I see to get a maximum of points is to populate an union of isochrone intersections. Going back to your example, we have to calculate the intersection of isochrone(Karlsruhe, 5 minutes) and isochrone(Pforzheim, 55 minutes), the intersection of isochrone(Karlsruhe, 10 minutes) and isochrone(Pforzheim, 50 minutes), etc. and then populate the union of these intersections. But I'm doubting in performance if we populate this union...

Finally I wonder if it's not faster to slice my map into rectangles, to calculate the distance matrix between all the centers of these rectangles and then identifying all the rectangles that match the condition. It may be faster but not easier because slicing the map is an hard work

What do you think about that ?

[Bernd Welter]

Could you provide an image how you'd like to slice the map?
I don't get the point

How about using xTour.findUnscheduledOrdersForTour?

http://xserver.ptvgroup.com/fileadmin/f ... s%7C_____8

[Bernd Welter]

And you can still calculate an airline ellipse to filter the candidates C for a later validation.
The (A:C)+(C:B) routings should then be performant enough...

[clegroux]

Here is a map example

Yeah good idea! I could first define the area with two isochrones and then calculate the "squared" ellipsochrone with (A:N) + (N:B) matrix requests.

I would like to ask you others questions about the distance matrix :

• 1- My map will be sliced into (approximatively) 6000 squares. How long will it take with xDima to calculate this matrix?
  2- According to your experience, could it be possible to define the "squared" ellipsochrone with (A:N) + (N:B) matrix requests with a 6000x6000 distance matrix?

Thanks a lot for all your posts Bernd