Set variable distance/time limits in Model Builder for Network Analysis

Question

Hey All

I'm using ArcMap 10.2.2. I am working on a routing project, it requires me to produce a network that is based on vehicle statistics. One of the desired functionality that is required is for this route to be within x% of the fastest and/or shortest route between the same points. I have developed a model that will create routes based on the shortest, fastest and one based on the statistics with no distance/time restrictions.

What I want to be able to do is pass the Total_Length (from shortest) and/or the Total_Time (from quickest) plus x% into the network based on vehicle statistics as a restriction. How do i go about doing this in Model Builder? I have tried playing around with variables but have had any luck.

Any help is appreciated.

Cheers in advance

Hamish

HamishKingsbury1 · Answer

Hi Michael

Thank you for that. I have spoken with some other team members and we have come to the conclusion that using a weighted combination would be the most appropriate approach for this project.

Thank you again, to both you and Melinda Morang‌

MichaelRice · Answer

Hi Hamish,

Unfortunately, the problem you are trying to solve is a variant of the notoriously-difficult “Resource Constrained Shortest Path Problem”. There are algorithms for solving this problem, but they can require exponential time to run in the worst case. So, basically, you cannot hope for a solution which is always practically-fast for all inputs of this problem type.

Furthermore, as Melinda pointed out, there is currently nothing in Network Analyst to support such constraints directly anyway. So, even if you wanted to use one of these algorithms, you would have to implement it yourself as a "custom solver".

Therefore, unless you want some help/advice on implementing a custom solver in Network Analyst from scratch (which is no trivial task), you will have to consider alternative formulations or be willing to relax your constraints. For example, assuming you have two separate cost attributes (let's call them A and B) , you could always define a third cost attribute (C) which represents some weighted combination of the two (e.g., C = x*A + y*B, for some x,y values in the range between 0.0 and 1.0). This would allow you some form of control over how you wish to weight the contribution of each individual attribute in the resulting paths if solving for the "hybrid" cost attribute C.

Let us know which direction you might like to take this in, and we can discuss any further options with you then.

HamishKingsbury1 · Answer

Thank you for that, hopefully someone has an idea of a way to approach this.