Latest Contributions by JaySandhu

The number one reason for not finding paths to all the destinations is network connectivity or some restriction that is causing the network to be disconnected. You can check if the network is connected in a couple of ways: 1. Use the Network Identify tool. It is on the Network Analyst toolbar. Second icon from the right. Activate it and then click on a network edge. It will pop up a list of the attributes of that edge and the edges and junctions it is connected to. Is it returning what you expect? Are any restriction attributes set that will cause the edge to be not traversed? 2. Use the Service Area solver. Load one of your origins from OD as a facility. Bring up the properties of SA layer and on the Analysis settings tab set a break value of, say, 10 minutes (if using time). On the Polygon Generation tab, turn off the polygons. On the Line Generation tab, turn on the generate lines. Close the properties and solve the service area layer. This should draw all the lines connected to that origin within 10 minutes. Do you see what you think you should see Or is there disconnects in the network? So if you do see disconnect, one potential thing to watch out for is the connectivity setting of the network dataset. Is this data supposed to connect at each edge end point OR is it supposed to connect to any vertices along an edge (yes if you got the data from a source like open streetmap). In that case you could switch the connectivity policy to Any vertex and rebuild the network. After a re-build of a network, make sure to, in the NA window, right-click the Origins and Destinations and recalculate location fields before you solve the layer again. Jay Sandhu ... View more

You can use Network Analyse with external LP solvers using PuLP. An example to do an allocation by first computing the travel cost matrix in ArcMap is here: http://www.arcgis.com/home/item.html?id=c5ae481f25604d1bb050d9bd1d6e3c06  Jay Sandhu ... View more

As in your previous post, you have described a process that has given you an unexpected result. It is probably because the iterative process you are choosing will not give you a global solution. Perhaps what you should state is the problem you are trying to solve rather than asking why your process did not work. ... View more

Sorry I do not understand what you are saying about what is shown in the figure you attached. Jay ... View more

All of network analysis is based on minimizing a "cost" on each edge. This does not have to be distance. It can be travel time or it can be travel cost. You just need to set up an attribute in the network dataset that is based on your cost. So yes, you can use a cost instead of distance. Location-Allocation first finds the minimum cost matrix between all facilities and demand locations and then minimize the weighted cost. The weighted cost is weight at the demand location times the cost to reach that demand location from the facility. You load in the weight for each demand location when you load in these locations into a LA demand point layer. The facilities too have a weight that can be loaded. BUT, the facilities weight is NOT used by the normal minimize weighted impedance (P-Median) solver. The facility weight is ONLY used by the Maximize Market share and Target Market share solvers. In those instances, the huff model or gravity model is used to determine the interactions between demand and facility (candidates and competitors) to determine the best location. ... View more

You can come up with a couple of solutions: 1. Once you have solved Location-Allocation once, you can set up a Closest Facility solver, load the chosen LA facilities as CF facilities, demand from LA as incidents in CF and solve for 2 closest. 2. You can solve Location-Allocation for the "first" facility. Then remove the chosen one and solve again. Now you will have the second facility. You will probably have to do this once for each chosen facility in your original solution. ... View more

Are your "orders" every household? Can you look at aggregating the orders per street or per side of street? That is, place one order per side of street and sum up the amount to pick up and service time to that order? You can run Add Locations on the orders and then Summarize the values based on the SourceOID field. And you can generate "mid points" to attach these values to with the Feature to Point tool (with the Inside option). Jay ... View more

I suspected that you were doing garbage truck routing. The VRP solver is currently not designed to do such high density routing. There are other ways to solve this problem where you need to visit each street (such as in postal deliver and snow plow routes) and the solution could be with an arc routing approach, sometimes also called Chinese Postman solvers. You can see some previous discussions here: https://community.esri.com/thread/6102#comment-402423 Let me add one suggestion. You have 7000 locations for 8 trucks. Which works out to be 875 per truck. Can all the trash be picked up by one truck without the need to empty it? That is, can you further subdivide the work based on capacity? ... View more

No need to get to me! I am here!! Can I ask the use case for so many orders per route? That is, we designed the VRP to optimize deliveries per day. And normally one truck or route can do about a 100 orders a day max.  And as you noticed, that works quite well. VRP (and TSP) is a combinatorial problem, so it will take longer and longer to solve for more orders per route. And with 875 orders per route it will take a while! Also note that when you solve for 7000 orders in one instance, it has to generate a complete OD cost matrix (with 49 million entries) and that takes cpu resources to compute and memory to keep that around while the VRP solver is clustering and optimally sequencing (TSP) the orders per route. So let me ask again what is the use case before I try to think of some alternative ways to tackle the problem. Jay Sandhu ... View more

Please do not add true "z" data from a DEM to a building data as it will destroy the existing artificial Z's one may have to make a building with different floors offset from ground. At a minimum, you need is a building line feature class with polylineZ for the hallway centerlines for you to route on and in addition, the stairs, elevator lines connected for the movement between the floors. You can have more as needed for restricted doors,etc. For visualization you may need other polygon feature classes to show offices, building, etc. Jay Sandhu ... View more

Lidia, You can have all the features in one feature class.You can use ArcGIS Pro's 3D editing capabilities to create and maintain such 3D building data. Jay Sandhu ... View more

As others have mentioned you can iterate with service area lines many times to find the break value that gives you 20 miles. Or you can use another approach: Location-Allocation. Let's say you have a field on your streets with length in miles. Run the GP tool Feature To Point with the Inside option on the streets. This will create a mid point on every street. Make a location-allocation layer. On the advanced settings, choose the problem type to be Maximize Capacitated Coverage. Load your facility location and set its capacity to 20. Load the mid points (or a relevant subset around your facility). Make sure to map the Length in miles as your Demand weight. Solve the location-allocation layer and it will select the points that sum up to 20 miles. There will be some "noise" at the ends as it may pick a further away point to ensure you get 20. But it will be quite close to the answer you are looking for. Location-allocation analysis—Help | ArcGIS for Desktop Jay Sandhu ... View more