Wireless - can u tell direction and/or distance to other laptops?

[Carl-Daniel Hailfinger]

Ivan Krstić wrote:
> John Marshall wrote:
>> Sorry if this has been asked already.  Will it be possible to tell
>> distance and/or direction from one laptop to another?  If so, any idea
>> how accurate such figures might be?
> [...]
> It'd be interesting, from a research point of view, to see how things
> would work if you distributed the problem, and had a number of laptops
> working together to calculate relative distances and such. It'd take
> some work, though.

If the measurements of other nodes are available, it should be possible
to get a reasonably accurate map (3-5%) with relative distances even
in the face of multipath propagation and a few obstables in the signal
path (note I wrote "a few", definitely not lots of buildings).

>> What I'd ideally like to know is which of distance, direction and
>> location may be achievable on the laptop and to what accuracy.  At this
>> time, however, I'm only looking for an rough indication of if it's
>> impossible, difficult, easy or simply unknown.
> 
> Some combination of impossible and difficult. There has been talk of
> throwing a GPS chip into the machines, but not for generation 1. In
> medium-noise environments where you've done previous data collection
> legwork, you can usually do wi-fi positioning within about 50m. Without
> previous data collection, you're pretty much out of luck.

How extensive does the data collection have to be? 50m should be
achievable outside of buildings with only two known fixed points, no
previous data collection, but distributed on-demand data collection.

> I actually researched wi-fi positioning in a previous life. The

Such an activity seems to be popular ;-)

> conclusion, from a paper which I never found time to get published:
> 
> "[...] we investigated the use of 802.11 access point
> beacons to determine location and group devices by location.
> Our experimental findings indicate that the high degree of
> precision demonstrated by Bahl and Padmanabhan [1] probably
> cannot be attained in heterogeneous, real-world systems. Our
> findings, although not conclusive, suggest that a median error of
> a few tens of meters, similar to that achieved by LaMarca et al [2].
> in Place Lab, is the highest level of accuracy achievable outside
> the laboratory. We do find, however, that beacons can be used
> effectively to match groups of users within several tens of
> meters of each other, which should be sufficient for most social
> networking applications."

How many entities did you collect data from to achieve your
results?

Actually, some of our problems may be solvable in a subset of
environments if:
* we know one fixed point (the antenna of the school server)
* children know the direction where that fixed point is
  ("where is the antenna on your school roof?")
* the signal strength for each unit has been measured by the
  school server when the unit was at a fixed predeterminded
  location
* no obstacles are between the school server antenna and
  each unit
* the signal strength measurements are reasonably accurate

Regards,
Carl-Daniel