Since they're already using the Google Maps API, they could try to
adapt the javascript-ified TSP solver:
http://gebweb.net/optimap/

Only works well for about 25 points though--but you could use
pre-sorting to divide it up into 4 possible best clusters, and then
apply the TSP solving on each of them.


On Sat, Oct 18, 2008 at 4:10 AM, Jennifer Strahan
<[EMAIL PROTECTED]> wrote:
> Hello Peter, Michael, Josh and Geowankers,
>
> Thanks for the detailed description of the problem and for the potential
> solutions.
>
> Does anyone know if there already exists a tool that canvassers can
> use?  Peter is correct that time is of the essence so there isn't any
> time to develop anything.  Or, are there any folks in the group willing
> to put in some volunteer time to quickly create something?
>
> Jennifer
>
> Peter Batty wrote:
>> This is actually a (somewhat constrained) traveling salesman problem
>> rather than requiring a point to point route, so Google Maps routing
>> won't help you. pgrouting supports traveling salesman but I haven't
>> used it and don't know what a practical number of points to handle is
>> - traveling salesman is a complex problem of course.  I was out
>> canvassing for the xxxxx campaign recently (the number of x's is a
>> clue, in my case at least :) !!), so was an end user of what I assume
>> to be the same system here. Of course I immediately wondered about a
>> better automated solution than they had also.
>>
>> I thought it was an interesting problem though so it's worth
>> explaining it in a little more detail based on my experience. We went
>> out in pairs, and were each given a package of paper sheets. The cover
>> page had a printed Google map with markers indicating the houses we
>> were to visit (this was the same map for both of us). Then behind that
>> we had a set of printed sheets, one or more per street, listing the
>> houses we needed to visit on each street in numerical order, with
>> details about the person/people to talk to at that house. We were just
>> visiting houses of known sporadic supporters and independents, so it
>> was a subset of houses - in this case it would vary from maybe 1 to 4
>> out of every 10 houses. One canvasser had odd numbers and one even
>> numbers, so you would do opposite sides of the street, so you had
>> someone in the same general area for support. Often you would have
>> more houses in a block on one side of the street than the other, so
>> one person would get ahead of the other. In the area that we were
>> canvassing, the blocks were long and thin, so we ended up walking
>> along the blocks "lengthwise" as most of the addresses we had were on
>> the north-south streets, but it was hard to tell if you were close to
>> a house on one of the east-west streets (on a different page from the
>> one that you were currently looking at). We ended up missing out some
>> of these. In total we had 90 houses in the package, 45 each.
>>
>> So what we really wanted was to each have a list of our 45 houses in a
>> suitable order for us to visit (as opposed to being listed street by
>> street in numerical order), with a map showing the route. Doing the
>> odd / even thing properly is an extra complication (i.e. taking 90
>> houses and coming up with two routes, which ideally keep the two
>> people close to each other). The simplest initial solution to this
>> would probably be to take all 90 houses, come up with the best route
>> to all those, and then just split it into odd and even after doing
>> that. If you got that working, then you could look at something cleverer.
>>
>> It seems as though for pgrouting you would need to have a reasonable
>> street network, which you may or may not have. In some cases,
>> especially if you had a pretty dense set of houses to visit, you might
>> be able to get a reasonable solution just using the locations of the
>> houses and ignoring streets altogether - but clearly in some cases
>> this would not work well.
>>
>> A pragmatic short term solution might be a semi-interactive one -
>> display the houses to visit on a map, let a user sketch a line
>> visually with an approximate route, buffer around that and find all
>> the houses close, and sort them appropriately based on that. And have
>> the ability to highlight any houses that were not yet added to the
>> route, etc. I suspect that for a short term solution (which is
>> presumably what you need), given the challenge of getting a good road
>> network, etc, this approach may be your best bet. It would need a bit
>> of custom development though, unless someone happens to have something
>> like that lying around.
>>
>> Cheers,
>>     Peter.
>>
>> On Thu, Oct 16, 2008 at 12:01 AM, Josh Livni
>> <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote:
>>
>>     pgrouting if you have the street data in postgis would be one way.
>>
>>     tho as long as they're drawing over google maps, why not insert a
>>     little javascript to use the gmaps api routing?
>>
>>       -josh
>>
>>
>>     On Thu, Oct 16, 2008 at 1:03 AM, Jennifer Strahan
>>     <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote:
>>
>>         Hello Geowankers,
>>
>>         I'm forwarding this email from a colleague in hopes that
>>         someone will
>>         have some suggestions to pass on.
>>
>>         Thanks for the help.
>>
>>         Regards,
>>         Jennifer
>>
>>         ps. I've stripped out political references.... that's why
>>         you'll see
>>         xxxxx campaign.
>>
>>         -------- Original Message --------
>>         Subject:        GIS routing question
>>         Date:   Wed, 15 Oct 2008 14:32:43 -0700 (PDT)
>>         From:   [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
>>         To:     [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
>>         CC:     [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>,
>>         [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
>>         References:
>>         <[EMAIL PROTECTED]
>>         <mailto:[EMAIL PROTECTED]>>
>>
>>
>>
>>         Hi All
>>
>>         My partner is working for the xxxxx campaign and was asked by
>>         the regional campaign office to look for computer-based
>>         routing solutions for canvassing.
>>
>>         Nationally, all the campaign offices use a networked,
>>         web-browser-accessible database called VoteBuilder. It manages
>>         contacts and lets field coordinators define canvassing
>>         territories by drawing over a Google map.
>>
>>         VoteBuilder doesn't construct a route for the canvassers--
>>         that's up to them. In suburban neighborhoods with winding
>>         roads and cul-de-sacs it's almost impossible for them to come
>>         up with an efficient route that doesn't miss some of the
>>         households.
>>
>>         Some of the local and regional offices have adopted software
>>         solutions for routing, but the techniques aren't being
>>         disseminated.
>>
>>         I know there's an extension for ArcGIS to do routing. Is this
>>         the only solution?
>>
>>         Thanks for any suggestions..
>>
>>         Louis
>>
>>
>>
>>
>>
>>         _______________________________________________
>>         Geowanking mailing list
>>         [email protected] <mailto:[email protected]>
>>         http://lists.burri.to/mailman/listinfo/geowanking
>>
>>
>>
>>     _______________________________________________
>>     Geowanking mailing list
>>     [email protected] <mailto:[email protected]>
>>     http://lists.burri.to/mailman/listinfo/geowanking
>>
>>
>>
>>
>> --
>> Peter Batty - President, Spatial Networking
>> W: +1 303 339 0957  M: +1 720 346 3954
>> Blog: http://geothought.blogspot.com
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> Geowanking mailing list
>> [email protected]
>> http://lists.burri.to/mailman/listinfo/geowanking
>
> _______________________________________________
> Geowanking mailing list
> [email protected]
> http://lists.burri.to/mailman/listinfo/geowanking
>
_______________________________________________
Geowanking mailing list
[email protected]
http://lists.burri.to/mailman/listinfo/geowanking

Reply via email to