Re: ISSUE-31 relative points?

On Fri, Jun 10, 2011 at 6:17 PM, Alex Hill <ahill@gatech.edu> wrote:
> I think we need to have a discussion about what "relative" positioning
> means.
> One example I have seen is in CityGML where building geometry is described
> in meters relative to an anchor point.
> However, I'm not sure what such a system for describing polygons, etc. would
> buy us in the POI spec.
> I tend to think that we intended relative positioning to facilitate things
> like the following:
> <pois>
> <poi id="frame_of_reference">
> <location>
> <!-- some reference to a moving vehicle a building or any arbitrary frame of
> reference (not just coords but orientation would be nice) -->
> </location>
> </poi>
> <poi id="alex">
> <location>
> <!-- some value calculated within that frame of reference (WiFi tracking,
> etc.)  in something not WGS84 -->
> <point srsName="urn:ogc:def:crs:EPSG:6.6:4979" srsDimensions="3">
> <pos>10023.123 1234.123 666.66</pos> <!-- lets say this is mm for now -->
> </point>
> </location>
> <relation type="relative-to" id="#frame_of_reference"/>
> </poi>
> </pois>
> Is that appropriate, or do we want something where each individual
> georeference can specify a relative frame of reference?
> <pois>
> <poi id="frame_of_reference">
> <location>
> <point relative_to="#frame_of_reference"
>  srsName="urn:ogc:def:crs:EPSG:6.6:4979" srsDimensions="3">
> <pos>10023.123 1234.123 666.66</pos> <!-- lets say this is mm for now -->
> </point>
> <!-- some reference to a moving vehicle a building or any arbitrary frame of
> reference (not just coords but oriented would be nice) -->
> </location>
> </poi>
> <poi id="alex">
> <location>
> <point relative_to="#frame_of_reference"
>  srsName="urn:ogc:def:crs:EPSG:6.6:4979">
> </location>
> </poi>
> </pois>
> Sorry if this is not "correct", but I'm more concerned about the spirit of
> this than the syntax.
> Would this be a valuable addition to the spec?

I really think so, I've always advocated relative positions in the
spec. I would actually love to see a fully hierarchical model, where
you can

- place POIs inside other POIs to make the relative relationship implicit
- have the applicable coordinate reference system cascade down until
explicitly changed

> Can we think of use cases that would benefit?

Indoor navigation/augmentation, obviously. It would be very
painstaking and error-prone to map out an entire building in WGS84
when you want to pinpoint all the coffee machines to centimeter
accuracy. Accurately pinpointing the front door once and offsetting in
millimeters from there is relatively (hah) trivial, especially with a
laser measure.

> I suspect the AR advocates can.
> What about indoor tracking of POIs?
> What about vehicle/pedestrian entrances for the mapping crowd?
> Can we put the relative information in the data model and then let systems
> back out the actual coordinates for efficiency?
> Then if you find (don't know how yet) that an update is necessary you can
> re-calculate.
> Perhaps what you have above ends up looking like:
> <location>
> <point>
> <pos>-72.123 34.1234 100.7</pos> <!-- the non-relative position of this
> point in WGS84 -->
> </point>
> <point relative_to="#frame_of_reference"
>  srsName="urn:ogc:def:crs:EPSG:6.6:4979" srsDimensions="3">
> <pos>10023.123 1234.123 666.66</pos> <!-- lets say this is mm for now -->
> </point>
> <!-- some reference to a moving vehicle a building or any arbitrary frame of
> reference (not just coords but oriented would be nice) -->
> </location>

What I'd like to see is that the applications calculate and cache the
derived coordinates (if they need it, that is. For display purposes,
locations need to be converted from whatever applicable CRS to display
coordinates anyway. Why would WGS84 be easier than meters?) If this
calculation is cumbersome (mobile clients) a POI hosting service might
do the calculation on the fly and serve it with the response or host a
separate service that does relative-to-absolute mapping. The further I
get into this paragraph, the less value I see in converting everything
to WGS84...

Jens

Received on Wednesday, 15 June 2011 07:58:23 UTC