VRP - MaxOrderCount Limt of 1,000?

[TimDolney]

I am using residential locations (homes) as locations for garbage collection.  The homes are the "orders" and I have 1 garbage truck for the "route".  I have 1,161 homes so I entered a MaxOrderCount of 1,161 for the garbage truck in the settings.  When I try to solve, I get a warning that the MaxOrderCount is greater than 1,000 and thus it won't solve.  Is 1,000 the max for the MaxOrderCount setting within Routes?

Thanks,

-Tim

[HeatherMoe]

Hi Adnan,

The VRP REST API runs off of the ArcGIS Online VRP service which is where the 200 max order count that you are seeing is imposed.  The 1000 max order count that was being referenced above is for using the ArcMap tool with a local network dataset or through the equivalent GP tool (not the ready to use service).  The GP tool can be published to create your own VRP service but would require an ArcGIS server with network analyst instead of just using service credits.  

More information can be found Overview of the Network Analyst geoprocessing service examples—Documentation | ArcGIS Enterprise 

[RachelApplebaum]

You could use some preprocessing to solve a kind of high density routing problem. It's not going to be exact, but it should be pretty good.

First create one point for each side of each street segment (where each point represents all the houses on that side of the street). That would prepare the data.

Then there would be 2 main options for the analysis:

1) Use the Location-Allocation solver to solve for as many facilities as there are routes, which will generate a cluster of stops for each route. Then run a separate optimized route for each cluster. Please note that this does not optimize the drive to/from the warehouse.

2) Use a Vehicle Routing Problem solver with dynamic seed points (See Route Seed Points class in Vehicle Routing problem classes < http://arcg.is/1XxDP25 >)

This won't change the MaxORderCount issue, but the results should still simulate what you're overall goal is.

[TylerPackett2]

Realy disappointed to find this is still an issue in 2020 in both ArcPro and ArcMap. Are there any plans to increase this limit so that we can solve common problems like trash collection and newspaper delivery with the tool that supposedly was made for them? 

[JaySandhu]

Tyler,

The VRP solver is designed for delivery scenarios where stops are spread over an area. It was not designed for arc-routing applications such as residential trash collection where every house on the block is being routed to. You can reduce the density of stops by aggregating the stops to one per edge or one per side of edge and use the VRP solver. Or you can use the Route solver to optimize one route at a time with more than a 1000 stops. 

What is your use case for a single route having more than a 1000 stops?

Jay Sandhu

[TylerPackett2]

We were evaluating the tool for a few things, one was for grassy curb strip spraying for areas off of roadways using the start, middle and end of blocks as the delivery locations. Only some areas have grassy curbs and there are multiple service vehicles. But even with that level of generalization, it is very easy for a truck to hit way more than 1000 stops before returning.

Other tasks that we would like to be able to use this for are asset status collection. Where we would use this tool to create a route for an auditor to follow and check on the status of assets using something like ArcGIS Quickcapture to collect more than 1000 locations per day per auditor. 

I understand increasing the number of stops greatly increases the complexity of the problem but our test system is able to solve 24,000 locations divided among 24 routes in less than 24 hours. Even if it's not advisable it would be nice to have the option to exceed the 1000 limit with a warning.

[JaySandhu]

Thanks for the additional details on what you are trying to do. In your test system, do you already have the 24,000 locations divided into 24 routes and are trying to optimize each route OR you are trying to assign the locations to 24 routes optimally and then optimize?

Jay Sandhu

[TylerPackett2]

The 24 routes are just to get around the 1000 stops per route limitation.

From there I would have to take the 24 routes and attempt to merge some of them into routes that would actually take up an entire day. Definitely less than ideal and the results are sub-optimal. 

My main point is there are legitimate VRP that have over 1000 stops per route and that a limit of 1000 stops seems arbitrary and antiquated. I fully understand that as we increase the number of stops we would see a substantial increase in the solve time for a route. But in many cases, these types of routes only need to be updated maybe once a quarter and I really wouldn't mind setting my own limit on the number of stops and letting this process on a server for a week to get an optimized route. Especially with services like AWS EC2 and Microsoft Azure where you can pay per hour to use high end compute VMs.

[JaySandhu]

Tyler,

VRP does two things: divide (assign) the orders between routes and then optimizes the route. This is an iterative and time consuming step. But in your case it seems you are not that concerned with partitioning the orders into routes as you have a general area you want to cover with one route.

I will suggest that you use ArcGIS Pro 2.4 and the Route solver with the Find Best sequence option. It does not have a limit on the orders per route and It will solve much larger problems more quickly. For example I solved a 2000 census tracts locations in LA in about three minutes and 6000 locations in about a half hour. All 24,000 may still be too much for one solve, but you could do four solves of 6000 each.

I will also suggest that you could reduce your input points by a third by just keeping the middle points and not the start and end points and use a curb approach right side of the vehicle option when loading your mid-points.

Jay Sandhu