Network Analyst Route Solver creates sub-optimal solution

[LorenzoFerrari]

Hi, I am trying to solve a route optimization problem with the Route solver in ArcGIS Pro (3.1). The problem is quite simple, and there are no complicated parameters or configurations involved.

When executed, the tool creates a solution that scolves the problem, but this solution is not the best. I can state this with certainty because by manually setting the sequence of stops to be executed instead of using the "Preserve first-last" option, the tool manages to solve the problem with a lower cost. Specifically, there is a precise area along the route that is not optimized. As seen in the screenshot, the proposed solution involves arriving from the right side of the image, passing over stops 14-15-16-17 (which are NOT executed now), continuing to stops 6-7-8-9-10-11-12-13, and then returning on the previously crossed road to execute stops 14-15-16-17. After executing these last 4 stops, the solution proposes turning at the roundabout (no U-turns at junctions) to get back and continuing south. 

The optimal solution would instead be to execute stops 14-15-16-17 on the first pass, so that you can immediately head south after executing stop 13.

The value for the Curb Approach parameter is "No U-Turns", so the stops should be executed on the first pass without considering the side of the street.

Any ideas?
Thank you

[JaySandhu]

It is hard to tell what could be going on without looking at all the point data. If you can share the data we can evaluate. You can zip and attach the stops file here.

Jay Sandhu

[LorenzoFerrari]

Hi, thank you for your help. Here's the stops file as requested. This feature class was saved after the tool execution. 
Please do not esitate to ask more information if you need. 
Thank you.

[JaySandhu]

What network dataset are you using to solve the route? I tried using ArcGIS Pro 3.1 and a version of Europe data from Here  from this year and I get different sequance with your stops. I got an even better result when I use the AGOL route solve which is based on Pro 3.3 and latest data. Attaching a screen shot from my Pro 3.1 solve.

Jay Sandhu

[LorenzoFerrari]

I'm using a small area of HERE data, probably the same dataset you used. I tried ArcGIS Pro 3.1, 3.2, 3.3 but none solved the problem. 1 thing i forgot to say (sorry) is that i'm solving the route using street traverse duration as impedance (a value in second i added for every street), not distance. Strangely, when solving with distance as impedence, the solver do much better and produce a solution that seems ok to me, similar to the one you posted. When solving with durating as impedence, the problem occurs: in that specific area, the route is sub-optimal.

Maybe i can give you the full network dataset i'm using, just to be sure we are analyzing the same set of data.
Import stops -> STOPS -> the Name, Sequence and CurbApproach attributes are presents for every stops.

Thank you again.

[JaySandhu]

Thanks for sharing the data. Will look at it.

Jay Sandhu

[JaySandhu · Accepted solution]

Lorenzo,

The python scripts for the cost/restriction evaluators had problems and that is why the time values being computed for the edges were not right. My colleague @MelindaMorang fixed the data and rebuild the network and now the route TSP sequence is better. I am attaching the updated network here.

Jay Sandhu

[LorenzoFerrari]

Thank you for your response, the proposed solution is now better on both 3.1 and 3.3 versions.