Considering no. of outlets that are to be visited by different vehicles (Vans, Bike, Truck) etc depending on the demand of the outlets & capacity of the vehicles, the routes generated in VRP have overlaps, keeping in mind that each vehicle should visit exactly 50 outlets. networkanalyst vrp geonet answers geonet #geo-visualization
I have tagged the network analyst group as well, but no response so far.
I have checked; I couldn't find anything on my query on the forum.
I was hoping if someone can help me, or need more details.
I'm not quite sure what you are asking. Are your routes overlapping (crisscrossing each other) and you are asking for how to make it so they don't overlap? Can you give a little more detail about what you are doing and what issue you are running into.
The issue is that the maximum no. of outlets served by each salesman is 50 in VRP.
Now when the VRP is executed, salesman one covers let's say 25 outlets from area one, another 10 from neighbouring area and rest from another. And 2nd salesman covers remaining 25 from area one and rest of the 25 from another and hence causing the overlap resulting in resource inefficiency.
You could try route zones for the different salesman. That would keep them within a region. The other option is to use seed points. That tends to give better clustered results without having to predesignate areas for the different salesman.
Hi Heather, I have used the route seed points. Depending on the nature of the data set, I couldn't always get clusters with no overlaps. We are talking about huge dataset that's random, for reference I can provide the sample data as well.
Even though route seed points give accurate results incase of generating routes for many salesman, but this doesn't hold true in all cases.
You have definitely hit on the short coming of any heuristic. It works well for a lot of different cases but isn't guaranteed to always do so. With the VRP solver using a heuristic that is also the case here and unfortunately it sounds like you have hit some of those cases where it is not performing as well as you would like.
A suggestion to try to help get more areas where the problem will solve satisfactorily with the VRP solver is to break it into smaller problems. The solver works better with cases under 1500 orders, but that said super high density is always a challenge for the solver since it does not do arc routing (routing to cover every street). Another option you could try to get better results is to pre-process the data so that orders on the same street segment are contracted into a "super" order and the number of real orders that it represents is shown using the pickup or delivery quantity.