Latest Contributions by MichaelRice

Bill, I noticed a similar posting by you in another thread here: http://forums.arcgis.com/threads/6551-Network-Analyst-Creating-a-Network-Dataset In that post, you said that you "downloaded the 3 file set from ESRI." Can you clarify what "file set" you are referring to? If you have Network Analyst enabled, you should be able to add any valid street network dataset to ArcMap and use it for your purposes. My initial suspicion would be that there is something wrong with the dataset you are trying to use (e.g., perhaps it is somehow incomplete, if you did not download/copy the dataset files correctly). ... View more

Can you provide some more detail on what you are asking? Perhaps including a simple example or some further discussion will better illustrate the problem you are trying to solve. Also, when you say "point", are you referring to a point that represents a stop in a route analysis layer, or are you referring to something else (e.g., a junction in your network dataset)? If you are just trying to determine the arrive time at a stop that is included in a route analysis layer, then you must specify a start time for your analysis layer. To do this, right-click your analysis layer in the ArcMap table of contents and select 'Properties...'. Select the 'Analysis Settings' tab on the properties dialog. Check the box for 'Use Start Time', specify the start time for your route, and click OK to close the dialog. Now, after you solve your route analysis, you can open the Stops attribute table and there should be a field called 'ArriveTime', which specifies the time of arrival for each stop in your Stops class. If this does not answer your question, please provide some further details on your problem, and I will attempt to provide a more appropriate answer. ... View more

It depends on what you mean by "nearest arterial road". If you mean "nearest" in terms of straight-line distance, then you do not need Network Analyst, and you can likely do this using a host of GP tools. If you mean "nearest" in terms of driving distance (or time) along the road network, then you would need Network Analyst. Using Network Analyst, you can create a new Closest Facility analysis layer and load your "points" (e.g., Point A, in your original post) as Incidents. Then load the vertices of your arterial road features as Facilities. Solve it to find the closest single Facility (i.e., the closest point on your arterial roads). Next, to get your path from a point to an urban center, create a Route analysis layer, and load your point as the first stop and your urban center as your second stop. Then solve to get the path. Finally, buffer this path as you wish. If this is all new to you, I would suggest you work through the Network Analyst tutorial, which should give you the information you need to get started. Let me know if you have any further questions or problems with the workflow suggested above (or if I somehow misunderstood you). ... View more

Zip files are certainly better, but the one you have uploaded is corrupted or invalid (try opening it to verify). There were problems at one point with attaching zip files to the forum postings; they were being corrupted. Perhaps this problem still exists. Please talk to the site administrators about this. I will see what I can do on my end as well. In the meantime, please let me know specifically what is confusing you about creating your network cost attribute, and I will try my best to clear this up. ... View more

I would be happy to help you out and look at your data, but your attachments appear to be invalid (or unknown) file types. You will have to give me more information on accessing your files or upload a more appropriate file type (e.g., a .zip file with your data). If you are having trouble uploading files, please contact the site administrator for help. ... View more

The "New->Network Dataset" option you mentioned is only enabled on a feature dataset if you have one or more valid source feature classes for your network dataset (e.g., a polyline feature class representing your streets) already in your feature dataset. Learning how to build your own network datasets should be covered as the first few exercises in the Network Analyst tutorial. ... View more

You have to add a network dataset to ArcMap before you can use any of the tools on the toolbar. I would encourage you to go through the Network Analyst tutorial that comes with your install of ArcGIS. This includes sample data, and will give you a good understanding of how to use the tools and work with the datasets. Please let me know if there are any further issues or questions. ... View more

Hi, I am glad to hear that you got past your original problem. As for your second problem of calculating travel time polygons for pedestrian travel times, your attachment appears to be invalid (i.e., I get an error message when I click on it). Please re-try attaching the image(s), and we can discuss this further. Regarding your last question about what fields must be present for your network dataset, there are several levels of detail you can establish in your data, depending upon your modeling requirements. The most basic requirements would be for you to have some source of travel costs for your edges. This is often represented as either travel distance (e.g., using a length field) or travel time (e.g., derived from length and speed limit fields) or perhaps even both (modeled as separate cost attributes in your network dataset). A cost attribute is required to perform any realistic shortest path analysis on your network. A next step would be to determine whether or not your data supports any restrictions for modeling real-world travel constraints on your road network (e.g., oneway streets or restricted turns). These can be modeled as restriction attributes in your network dataset. Oneway restrictions are often derived from simple boolean fields in your dataset (to indicate whether or not a street is oneway for a given travel direction). Restricted turns (or even turn delays) can either be modeled globally (i.e., applied universally to all possible two-part turns) or established for only an explicit set of turns in your network (this latter possibility requires you to create a turn source feature class). Finally, you might also wish to establish support for generating driving directions on your dataset. This requires a street name field, as well as a specification of a length cost attribute and time cost attribute. Of course, there are several other possibilities you might wish to consider when building your network datasets, but these are the most common. For a more detailed discussion of all of this (and more), including walkthroughs of building network datasets on sample data, I would strongly encourage you to work through the Network Analyst tutorial (this comes with your installation of ArcGIS). This includes examples of everything I have touched on above, as well as some more complex modeling scenarios. This is the best place to get started, and this will help you get a feel for how to setup your own network datasets using your own data. If you have any further questions, please let me know. ... View more

Here is a quick checklist to help troubleshoot this issue: When you made sure the extension was checked on, did you do this from ArcMap or ArcCatalog? If you are trying to build the dataset in ArcCatalog, you must ensure the extension is enabled there. Ideally, if you are using Network Analyst, you should enable the extension in both ArcMap and ArcCatalog. Is your source polyline feature class really a shapefile, or is it perhaps a source feature class in a geodatabase? If it is a shapefile, then you should be able to right click on it to create a new network dataset, as you suggested. However, if it is a feature class in a geodatabase, then you must right click on the feature dataset that contains the source feature class to create a new network dataset. If neither of these possibilities helps to resolve this issue, can you please post some screenshots showing that you have enabled the extension in ArcCatalog and that the "New Network Dataset" option is still disabled? ... View more

Well, it's not quite as simple as you seem to suggest... Bill, My point was that, if you are willing to accept some approximation to your formulation using my original suggestion (i.e., considering only vertices), then you could eliminate some of the negative impacts of having "long-distance" edges by splitting them up before-hand (and doing this only once, before running the originally-suggested algorithm). This is clearly not an optimal solution, nor was it suggested as such; perhaps it was not what you were originally hinting at, but it is indeed quite simple. This suggestion could be done using our existing out-of-the-box software. Any dynamic algorithm, such as you suggested, while nice, would require the development of a custom solver. ... View more

Jamie, This is quite interesting. This could be formulated as a classic location-allocation problem type. That is, given a set of k candidate locations to choose from, decide which j (< k) locations (in your case, j=1) are best suited to "service" a set of demand points, such that each demand point gets allocated to its nearest chosen location. There are several ways to formulate the objective function for solving such a problem (e.g., minimize the total weighted impedance to the demand locations, maximize the number of demand locations which are within a specified coverage impedance, etc.). So far we have been discussing this problem in the context of finding the single junction (the "service" location) which minimizes the maximum impedance to all other junctions (the "demand" locations). However, we could easily reformulate the problem to choose the junction which minimizes the total (and thus, average) impedance to all other junctions. This helps to eliminate the effects of having one or two outlier demand locations. The point is that there are many available options for defining your objective function for this problem type. We have developed a new solver in Network Analyst at the 10 release to solve just this problem type: the Location-Allocation solver. Its capabilities are much more advanced than I have even attempted to describe here, with many more options for greater flexibility in defining your objective function for selecting/locating your facilities. I would imagine if your tri-county region is serious about establishing a proper location for their distribution center, then they have to consider the possibility that they might not be able to simply establish such a facility at any possible junction in the network (e.g., perhaps there is no available real estate in a given area of your network coverage). Therefore, using the Location-Allocation solver, they could effectively specify which candidate locations (even ones located midway along edges) that they are willing (and able) to select from, then specify the number of the distribution centers they are planning to establish (e.g., 1, so far in our discussion), and then specify exactly (and only) those locations which the distribution center would truly be servicing (this may or may not include all junctions of the network; there may be a small number of explicit locations to be serviced by this distribution center). Please stop by the island at the UC and we can discuss the options in further detail. ... View more

Bill, Our respective formulations of this problem are merely two slightly-different interpretations of the same general property. The goal (as I understood it) was to determine a single position on a graph, from which you can reach all other positions in the graph in <= x units, where x is the minimum possible value. When said "positions" are limited to vertices, this is exactly what my formulation provides (based on the well-known concept of graph radius), and it applies to all graphs (even your example triangle graph). In your formulation, it appears that you have merely extended this concept of "position" to apply to both vertices and edges (i.e., a "position" can mean any position along an edge as well). As we have both pointed out, this latter formulation is a bit trickier. However, as you also suggested, this latter formulation can still be approximated using the original concept of graph radius (which applies only to vertices) by simply introducing new dummy vertices along any "long-distance" edges (and thus, "splitting" the edge into shorter constituent edges). So, the good news is, at the very least, it appears that we have some general understanding of the various possible formulations of the originally-suggested problem. I will leave it to Jamie (the originator of this thread) to decide which is more suitable for their purposes, and we can go from there. Jamie, can you perhaps tell us why it is that you wish to determine this "centroid" position in the first place? Depending upon your actual use case, there may be other, simpler ways of achieving your goal. ... View more

I will be at the UC as well, staffing the Spatial Analysis Island on Tuesday through Thursday of next week. Please feel free to stop by and talk with me (or anyone else on the Network Analyst team), if you get a chance. I am interested in better understanding your use case for discovering such graph centers, and maybe figuring out how we can better support this case. ... View more

In order to calculate such a position on a road network (which is represented as a graph), you would ultimately be looking for the position with the smallest graph eccentricity (see http://mathworld.wolfram.com/GraphEccentricity.html for a description of this concept), which defines the radius of the graph. The only out-of-the-box way to do this in Network Analyst would be the following: Create an OD Cost Matrix analysis layer (set your impedance attribute to either a time- or distance-based cost attribute, depending on whether or not you wish to find the time- or distance-based center, respectively) Load all of your network junctions as both Origins and Destinations in your analysis layer Solve the OD instance (this essentially calculates an all-pairs shortest paths problem) For each origin, take the maximum cost to reach any destination; select the origin with the minimum such value for this maximum destination travel cost; this origin defines the center based on the graph radius Of course, this methodology will only provide you with the junction which has the smallest eccentricity. If you are looking for a more general position along the graph (e.g., potentially somewhere along an edge), then the process would be more complicated. Also, this method will not scale very well for very large graphs (with more than approximately 5000 junctions), due primarily to memory limitations for solving such large OD instances. How big is the network you are planning on computing this value for (in terms of the number of junctions and edges)? ... View more

Jeremy, Were you ever able to resolve your problem? ... View more