[Talk-GB] Adding a further 250, 000 UK roads quickly using a Bot?

SK53 on OSM SK53_osm at yahoo.co.uk
Thu Feb 3 11:03:06 GMT 2011

On 02/02/2011 21:10, Peter Miller wrote:
> ITO have been offering a service to compare osm road names with os 
> locator road names for a while now[1]  which has encouraged a lot of 
> activity - and has even led to Andy to obsession.[2] I have also 
> suffered from a bout of urgent mapping myself while completing all of 
> Suffolk to 95% in the past few weeks! Can I suggest that for our 
> sanity we should consider developing a bot to do some of this work for 
> us? This would also allow us to get the rest of the 250,000 remaining 
> roads in place in less than the 13 months Andy estimates will be required?
> This bot would do a number of repetitive tasks for us within the 
> bounding box in which it was authorised to operate by a contributor.
> It could do the following:
> 1) Add names to existing roads in osm where there is a single un-named 
> ways in osm with a bounding box which matches that of a single entry 
> in os locator.
> 2) In addition...  it might be able to also add roads to osm from os 
> vector district, snapping them into existing roads as required where 
> the existing roads align neatly with os streetview. It would only do 
> this if there were no ways close by on either side.
> Complex situations will be left to humans. Humans could also sometimes 
> prepare an area for analysis by the bot, splitting ways as 
> appropriate, adjusting alignment of existing roads and dealing in 
> advance with situations we know the bot will have difficulties with.
> Edits would be made as individual changesets, referenced to the mapper 
> operating of the bot. Each edit would be 'signed off' by the mapper 
> who would be able to see the proposed changes visual prior to 
> accepting them.
> Any thoughts?
> Peter
> [1] http://www.itoworld.com/product/data/osm_analysis/summary
> [2] 
> http://www.gravitystorm.co.uk/shine/archives/2011/02/02/the-london-streets-challenge/
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-gb

I personally don't think it's a great idea. There are many aspects to 
this, so I'll just take the ones which occur to me right now.

    * Automating road completion is not a huge impossiblity. In a
      desultory way I have been playing with name assignment to
      VectorMap District data, and I'm sure that approached in a more
      systematic and determined fashion it would feasible to produce a
      programmatic way to assign Locator names to the VDM data set. I
      have a pretty good idea of the major issues, and the outline
      algorithms to do this. The main step I have not tried is sticking
      combinations of linestrings together to maximise fit to a locator
      box. Connectivity is the other main issue. However, I think such
      data would be pretty much useful on their own, or could be
      mashed-up with OSM data for a given application (e.g., a garmin
      map). I don't see huge immediate utility in putting such data into
      OSM, as opposed to making it available in such a way that it could
      be integrated with OSM data.
    * Relying on OS data reduces the range of sources and validation of
      OSM. I'm currently experimenting
      with the Nottingham area to see how close I can get to the Urban
      Atlas mapping done for the EEA. My gut feel is that OSM data which
      is sourced through combinations of ground-survey and aerial
      imagery can provide similar levels of accuracy in terms of areas,
      and higher levels of reliability in terms of landuse
      classification. In other words, once OSM data start to get very
      detailed they provide a separately surveyed source of data which
      is distinct from OSGB. If we populate huge swathes of the country
      from OSGB data we lose this value (but see next).
    * Huge swathes of the country *are to all intents and purposes* only
      populated with OSGB data. Most towns in Northern England appear to
      have been largely traced: obvious examples: Darlington,
      Middlesborough, Bolton, Oldham, Rochdale, Grimsby, many parts of
      South Yorkshire. In many cases tracers using OS StreetView have
      not bothered to add road names (from time to time, I have been
      doing sweeps through Oldham fixing some of these). The few
      journeys I have made in these places indicate a poor level of
      quality. Adding data does not add mappers.
    * Many areas traced from Yahoo have received little or no ground
      survey. Most of Merseyside, Greater Manchester and West London
      fall into these categories. In some cases a concerted effort has
      been made to at least add street names from OS StreetView. See
      comment above.
    * We are still nowhere near 1 dedicated mapper per City/District
      which seems to be the minimum to give a mapped area 'life'. Of
      course we really want a group of people maintaining an area, but
      I'd settle for A lifeless but complete map will not IMO attract
      these individuals.
    * Progress has been substantially affected by license FUD. This is
      not just about using OS OpenData, but the prospect that edits may
      be lost because an earlier editor in an area is no longer
      contactable or refuses the new terms. You, yourself, have
      expressed concern about this issue [1] <#1>,[2] <#2>. I anticipate
      the ODbL, CT issues will continue to affect how and what people
      map until the whole process is resolved. Pushing in a whole load
      of additional data won't help, particularly as strong resistance
      to ODbL may result in certain areas becoming rather sparse.
      Consideration of doing this after complete adoption of ODbL may be
      a different issue.
    * Availability of OS OpenData has opened whole new avenues for using
      data for many OSM contributors. Exploring these avenues have
      reduced time and effort devoted directly to generating OSM data.
      However, I would think that these explorations will produce more
      people able to manipulate OS and OSM data in more sophisticated
      ways, which in turn will result in richer OSM data.
    * OS OpenData is out-of-date. The April 2010 StreetView tiles are at
      least 2 years old, and where I've checked VDM is similarly dated.
      I have not failed to find a significant change between OS OpenData
      (and Bing imagery) in detailed surveys I've done this year. Chris
      Hill has a similar experience

So in summary, I have no problems with collectively working on an 
*external *enhanced data set sourced from OS OpenData designed to 
complement what is in OSM, and tools to link it with OSM data outside 
the planet db. Importing such data on a large scale will just leave us 
with a lifeless data set and a stagnant contributor community.

Jerry Clough

P.S. I long ago generated an OSM file with all Nottingham roads so as to 
be able to preemptively 'flood' any unmapped areas should mass tracing 
or import occur. I did this to avoid a reoccurrence of mass realigning 
of ground surveyed roads to OS StreetView as happened in the Carlton & 
Gedling area. The changes required to unroll this are too complex, so 
effectively this area requires re-mapping from scratch.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20110203/1969f920/attachment-0001.html>

More information about the Talk-GB mailing list