Hi!

I'm trying to build a triangle distane mesh for a big number of points in the map (in C# and using mappoint 2009). With 200 points, it takes about 2,5 hours. The problem is that I'd like to build a mesh of about 2000 points, that could take some days.
Is there any trick or some way to speed up the calc of the mesh?
If not, any alternative?

Best regards.

I am not quite sure what you mean by a mesh? Do you mean a physical net of lines?

Or simply an array of route calculations?
eg. all route distances from Group A to Group B (or A & B could be the same group).
If so, our MileCharter add-in can do this, with the output to Excel.
Thousands of routes are always going to take a long time - all routes from 1000 points to 1000 points is a million routes - this will take days to compute.

There are ways of speeding things up - eg. use straight lines, restrict the points used, etc?

Richard

Hi Winwaed,

I mean an array of route calculations. I'm trying MileCharter add-in and it works great, but seems that it's going to take long time too to calc 2000 points.
It seems that there is not any shortcut to have it done in less time

Regards.

Originally Posted by Winwaed

I am not quite sure what you mean by a mesh? Do you mean a physical net of lines?

Or simply an array of route calculations?
eg. all route distances from Group A to Group B (or A & B could be the same group).
If so, our MileCharter add-in can do this, with the output to Excel.
Thousands of routes are always going to take a long time - all routes from 1000 points to 1000 points is a million routes - this will take days to compute.

There are ways of speeding things up - eg. use straight lines, restrict the points used, etc?

Richard

I was wandering if is there any way to take adventage of using multiple processors to run the calcs in parallell using mappoint?
Is it possible to run an instance of mappoint in a separate processor using his own copy of the maps to really parallellice the calcs?

This is how MPMileage performs speedups. Note that you won't see much of a speedup with a dual core processor because MapPoint's routing is partially multi-threaded already. A quad core should show good speedups though.

MPMileage requires more setup. It works through a database that lists the requested from/to routes, and calculates them. If you do use MPMileage, make sure you read the docs, tutorial ,etc.


2000 routes shouldn't take *that* long - 40mins or so on a slow PC?


Richard

i think that is the general problem with network analysis isnt it.

you cannot completely "solve"/"optimize" a network analysis because its just too big


I forget all the correct terms, but isnt this to do with networks/graphs (mathematical terminology) etc.
The empirical method is to find a reasonable solution, and then see if you can improve it, within certain constraints.

Travelling Salesman Problem comes up a lot and NP-Complete Problems.

Or in English "Computationally Difficult" (the problem gets more difficult as the size increases in an exponential manner)


Richard

thinking again -

the best way is possibly to compute straight line distances between your points, rather than road distances.

that would be do-able and trivial. seconds, minutes at most