Page 1 of 1

Geographical grouping - PTV xServer 2 - xTour

Posted: Mon Apr 07, 2025 8:26 am
by jerich
Hi Bernd,

I am looking into optimizing several orders for CZ.

Here is a quick overview of 5 customers, the stop symbol is the 5th customer:
map.png
There is a special requirement: Customers which are located near each other (same postal code ares) should be planned together.
In this example: Postal codes starting with 2 is one group, starting with 7 is another group etc.

There should be no tour driving from east to west inside the country. I saw your forum post viewtopic.php?t=1859
optimal planClusters.png
I am curious: Is clustering something that can be used together with xTour? Or do you have another idea how to achive this?

Attached is the request and response that I am currently getting.
The response gives me 3 tours.
Tour 1 is a FTL this is ok.
Tour 3 is not ok - driving from east to west - it looks like this:
Screenshot_1.png
Screenshot_1.png (46.54 KiB) Viewed 4357 times
I already tried setting AvoidIntersections to true but this does not help here.

Re: Geographical grouping - PTV xServer 2 - xTour

Posted: Mon Apr 07, 2025 10:01 am
by Bernd Welter
Hi Matthias,

the main purpose of the cluster component is
  • to determine the groups from a strategic level based on "fair distribution within the clusters" (based on huge workloads)
  • and then use the additional group IDs in multiple, distinct operational plannings (e.g. during the next business year). Check this story as an example.
Here's some more thoughts about this:
  • The separation of strategic and operational planning used to be the solution for planning stories with a complexity where the operational components were not able to handle the big scenario. SInce Optiflow is able to handle many thousand orders in a single planning the need for clustering has at least decreased: some customers who were used to perform the 2 step approach (xCluster2+xTour2) are now solving their daily tour plannings with Optiflow in a single step.
  • If you apply a "zip code based" grouping this is understandeable at first - but doesn't reflect the absolute workload in the virtual areas.
From my perspective I'd say we would have to look into your single (small sized) tour optimization call to identify whether we can push the optimization to deliver the desired output.

I need some time to look into this, so please stay patient.

Bernd
animated result
animated result