[android-discuss] Re: (D)GPS good to 10 cm

2008-07-09 Thread whitemice

Having an accelerometer on a chip is a great idea.  They are cheap,
accurate and available on more and more phones:
http://www.dimensionengineering.com/accelerometers.htm

The problem is that you are trying to use them for inertial
navigation, which is inherently complex, requires both orientation and
acceleration sensing and some sort of position initialisation.  Don’t
forget that mobile software developers have to make do with *standard*
hardware.
http://en.wikipedia.org/wiki/Inertial_navigation_system

I see that you are clutching at straws, trying to make your use case
work.  To that end, have you considered investigating Wi-Fi based
positioning?
http://www.spotigo.com/products-and-services/spotigo-wifi-based-positioning-solution/

Bluetooth can also be useful, but I won’t comment on that until I have
been allowed to play with the actual Android API. ;-)

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-09 Thread whitemice

There are a lot of interesting use cases for sub 1m accuracy and/or
indoor positioning on a mobile phone – hence I am following your ideas
with interest.

Your proposal as I understand it:
If the communication API (at either the server or the client side)
gives you access to signal strength and packet delay information, you
may be able to get some idea of distance (but not direction?).  This
estimate would become less reliable as more walls and furniture get in
the path of the signal.

Communicating with two Wi-Fi transmitters (if this is possible on the
client side?) with known positions and coverage areas, would give you
two or more overlapping regions where the handset might currently be
located.

By fixing your results along known walking routes and integrating path
data (from the accelerometer) you might be able to guestimate a user’s
current position.

With all these uncertainties, I think claims of sub 1m accuracy are
overly optimistic (but not unachievable), although such a system could
have other benefits.

To the original question: the Android Wi-Fi API has not yet been
released, so I can’t comment on how well your proposal might work on a
real device.

Also, I am also investigating using accelerometers for estimating
position *inaccuracy* for my Snowball platform, which has
significantly lower accuracy requirements:
http://blog.zedray.com/snowball/

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-09 Thread [EMAIL PROTECTED]

whitemice,

Your description is technically in focus in that we are looking at the
same thing, I just want phone to use the accelerometer(s) to emulate 1
meter GPS and compass functions when no GPS or compass is available.

1.) So all the programming involves is: behavior is
f(dx,dy,dz,orientation). The system provides the data from whatever
costs the least. The function is already there on location, but
accelerometer (inertial) estimates were not on the list of location
service providers. This should be on the list with whatever caveats
come with it (like requires consistent orientation since last known
good).

2.) The other thing is that when data is bad and tower location cannot
resolve exactly where the phones are I want the system to present
choices  in addition to precision estimates. The local program can
have stored information about routes and history to  pick from several
possible positions.  This will maintain continuity in the application
whereas having precision suddenly go to a km with no more bearing data
really blows up a pedestrian level program that requires a simulated
compass.

3). Also, when comparing two poor signals (think like interpolating
steam tables), the comparison needs to be done before any rounding. I
want the phones to be able to share and compare raw data with each
other without asking the system to make an independent (non-
differential) estimate of where each is. If I am trying to locate
someone and I cannot read a map, all I care about is relative
position.  The system can act like some systems engineer and insist on
giving me only absolute data and throw the baby out with the
bathwater, but it does not have to.

In other words, android location services can be more robust than
systems costing 4 times as much.  This is a difference all customers
will recognize when they do the inevitable side by side comparisons.
Each time that happens Android takes share.

ed (the systems engineer :-)

On Jul 9, 7:05 am, whitemice [EMAIL PROTECTED] wrote:
 There are a lot of interesting use cases for sub 1m accuracy and/or
 indoor positioning on a mobile phone – hence I am following your ideas
 with interest.

 Your proposal as I understand it:
 If the communication API (at either the server or the client side)
 gives you access to signal strength and packet delay information, you
 may be able to get some idea of distance (but not direction?).  This
 estimate would become less reliable as more walls and furniture get in
 the path of the signal.

 Communicating with two Wi-Fi transmitters (if this is possible on the
 client side?) with known positions and coverage areas, would give you
 two or more overlapping regions where the handset might currently be
 located.

 By fixing your results along known walking routes and integrating path
 data (from the accelerometer) you might be able to guestimate a user’s
 current position.

 With all these uncertainties, I think claims of sub 1m accuracy are
 overly optimistic (but not unachievable), although such a system could
 have other benefits.

 To the original question: the Android Wi-Fi API has not yet been
 released, so I can’t comment on how well your proposal might work on a
 real device.

 Also, I am also investigating using accelerometers for estimating
 position *inaccuracy* for my Snowball platform, which has
 significantly lower accuracy requirements:http://blog.zedray.com/snowball/
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-09 Thread whitemice

(1) I’ve never seen someone trying to do inertial navigation on a
phone, but it could work temporally if you have 3 axis orientation and
acceleration sensitivity, or are only handling a 2D problem domain.

I just checked and unfortunately it would require a lot of messing
around to get JavaME access on my N95 to try this idea out.  So I will
wait for my first Android phone before experimenting with this idea.

(2) Agreed, multiple positioning technologies make the solution more
robust.

(3) Agreed, but wait for the API to see what’s possible.  Google have
so far been very quiet with important technical details like this.

What’s your background, are you going to try and build this?

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-09 Thread [EMAIL PROTECTED]

...are you going to try and build this?

Only if we have to.

We are going to use the expensive handsets (that have accelerometers
and compasses) as prototypes to show an end to end people coordination
tool. On each step the use model will be made to work well using only
those UI elements that will exist on the target hardware, a $50
handset.

Then it is about persuading a handset manufacturer to make a
pedestrian phone for ordinary people.

Most of the early effort will be centered around making carpooled
commutes efficient and attractive. The commutes will work with phones
with no compass or accelerometer. Those instruments really come into
play for queuing and organizing people on foot - which is one of the
reasons carpools don't work right now.

So yes, for an end to end tool the pedestrian stuff has to work. For
pedestrian stuff to work you need reliable GPS like differential
location with compass function.

Inertial bridging to help a handset application select from location
choices provided by a not quite sure tower triangulation service (an
extension of this service where GPS coordinates are correlated to
other parameters that phones know:
http://googlemobile.blogspot.com/2008/06/google-enables-location-aware.html
)
goes a long way toward making a phone that works for people on foot.


On Jul 9, 11:03 am, whitemice [EMAIL PROTECTED] wrote:
 (1) I’ve never seen someone trying to do inertial navigation on a
 phone, but it could work temporally if you have 3 axis orientation and
 acceleration sensitivity, or are only handling a 2D problem domain.

 I just checked and unfortunately it would require a lot of messing
 around to get JavaME access on my N95 to try this idea out.  So I will
 wait for my first Android phone before experimenting with this idea.

 (2) Agreed, multiple positioning technologies make the solution more
 robust.

 (3) Agreed, but wait for the API to see what’s possible.  Google have
 so far been very quiet with important technical details like this.

 What’s your background, are you going to try and build this?
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-08 Thread [EMAIL PROTECTED]

Am i really that illiterate?  Firefox is better (-:

On Jul 8, 10:36 am, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote:
 Thank you for the AGPS link.
 I agree that Cell ID is the right granularity for a lot of useful
 solutions. And appreciate the 10cm precision instrument size.

 I have a friend with a Garmin wrist watch sized GPS that resolves to
 about 3 ft by looking at the data from lunch run departures and
 returns on the same sidewalk.  That is good enough to have an item
 beacon.

 The accelerometer can theoretically simulate position data as you walk
 around in a store. If you calibrate your cart on a yellow arrow at the
 entrance... three radios in the corners of the store ceiling could
 give you 'particle cell' tower position data. Maybe at a low enough
 frequency to not get blocked by buildings.

 So I guess the idea is: An item beacon can be implemented in several
 ways, 1) an expensive, 2+ dollar, accelerometer that integrates twice
 to get accumulated displacement (if mounted to a cart that might
 actually work), and 2) event or location based radio system on the
 smaller than micro scale that help a market organizer provide guidance
 to customers at trade shows or other marketplaces.

 I don’t' know if you have ever struggled to find a booth at SuperComm
 or NextComm or whatever it is. They could get more money for their
 back alley booths if folks could choose an itinerary or list of
 destinations from the web and let micro radios in the corner locate
 and android phones guide them to booth by booth by presenting merchant
 beacons in a logical order. The calibration spot would be for
 promotional purposed mostly.

 Does anyone have an accelerometer to GPS utility for indoor mapping
 yet?
  Is the i-phone accelerometer good enough to do this?

 On Jul 8, 9:22 am, whitemice [EMAIL PROTECTED] wrote:



  Android will be restricted by the limitations of the hardware on which
  it runs.  Right now 10cm is achievable with a device the size of a
  toolbox, which can average out multiple input signals over time.

  You are more likely to find an implementation of “Assisted GPS” on an
  Android device:http://en.wikipedia.org/wiki/Assisted_GPS

  To get an idea of performance, have a go on a Nokia N95 - You will
  find that:
  - It doesn’t work indoors (shooting down the in-store use case).
  - Cold start times can be excessively long.
  - Detrimental to power consumption (if in constant use)

  Performance on handsets is still inferior to dedicated GPS devices.
  Don’t expect massive improvements in the near future as Mores law
  doesn’t apply to mobile devices.

  I blogged this in more detail here (half way down the 
  post):http://blog.zedray.com/2008/05/16/android-developer-challenge-critique/-
   Hide quoted text -

 - Show quoted text -
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-08 Thread [EMAIL PROTECTED]

This one needs a little clarification.
Kart racers have been using accelerometer based data acquisition
systems with trackside beacons for accurate lap times for a while
now.  The newer ones have GPS, but I don't think they always did.  The
granularity of the data is really fine. The instruments retail for
between $300 and $700 and sell in small quantities compared to,
say,... cellular phones.  With micromachining the accelerometer does
not have to be expensive to be remarkably good. Anyway for a sense of
the progression from accelerometers to GPS, read down to the bottom of
this page:

http://www.edgekart.com/store/gauges/mychron4.htm

Here is another supplier: http://www.digatronusa.com/products-karts.shtml

An accelerometer will make a scribble with no reference to the track
geometry, but with decent calibration, at a store entrance, an
accelerometer emulating a GPS can find item beacons on store shelves
(note: the phone behaves as if the item is sending a beacon based on a
coordinate difference between phone and item and an emulated compass)
even when no GPS [signal] is present.

This works for vendor booths at trade shows too, but the stable
platform of a grocery cart may be needed for early versions just
because signal to noise is so good compared to being on a person.


On Jul 8, 10:36 am, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote:
 Thank you for the AGPS link.
 I agree that Cell ID is the right granularity for a lot of useful
 solutions. And appreciate the 10cm precision instrument size.

 I have a friend with a Garmin wrist watch sized GPS that resolves to
 about 3 ft by looking at the data from lunch run departures and
 returns on the same sidewalk.  That is good enough to have an item
 beacon.

 The accelerometer can theoretically simulate position data as you walk
 around in a store. If you calibrate your cart on a yellow arrow at the
 entrance... three radios in the corners of the store ceiling could
 give you 'particle cell' tower position data. Maybe at a low enough
 frequency to not get blocked by buildings.

 So I guess the idea is: An item beacon can be implemented in several
 ways, 1) an expensive, 2+ dollar, accelerometer that integrates twice
 to get accumulated displacement (if mounted to a cart that might
 actually work), and 2) event or location based radio system on the
 smaller than micro scale that help a market organizer provide guidance
 to customers at trade shows or other marketplaces.

 I don’t' know if you have ever struggled to find a booth at SuperComm
 or NextComm or whatever it is. They could get more money for their
 back alley booths if folks could choose an itinerary or list of
 destinations from the web and let micro radios in the corner locate
 and android phones guide them to booth by booth by presenting merchant
 beacons in a logical order. The calibration spot would be for
 promotional purposed mostly.

 Does anyone have an accelerometer to GPS utility for indoor mapping
 yet?
  Is the i-phone accelerometer good enough to do this?

 On Jul 8, 9:22 am, whitemice [EMAIL PROTECTED] wrote:



  Android will be restricted by the limitations of the hardware on which
  it runs.  Right now 10cm is achievable with a device the size of a
  toolbox, which can average out multiple input signals over time.

  You are more likely to find an implementation of “Assisted GPS” on an
  Android device:http://en.wikipedia.org/wiki/Assisted_GPS

  To get an idea of performance, have a go on a Nokia N95 - You will
  find that:
  - It doesn’t work indoors (shooting down the in-store use case).
  - Cold start times can be excessively long.
  - Detrimental to power consumption (if in constant use)

  Performance on handsets is still inferior to dedicated GPS devices.
  Don’t expect massive improvements in the near future as Mores law
  doesn’t apply to mobile devices.

  I blogged this in more detail here (half way down the 
  post):http://blog.zedray.com/2008/05/16/android-developer-challenge-critique/-
   Hide quoted text -

 - Show quoted text -
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-08 Thread JP



On Jul 8, 4:20 pm, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote:
 This one needs a little clarification.
 Kart racers have been using accelerometer based data acquisition
 systems with trackside beacons for accurate lap times for a while
 now.
Have you driven a kart? These things GO. This means accelerometers
have plenty to work with. A person pushing a shopping cart around a
shop is probably not even raising sensor input levels over noise.


--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---



[android-discuss] Re: (D)GPS good to 10 cm

2008-07-08 Thread Muthu Ramadoss
What's the noise? If the accelerometer cannot detect a device moving at a
walking pace, then what's the real use for it? I'm not an expert, so asking
the experts here :)

On Wed, Jul 9, 2008 at 9:00 AM, JP [EMAIL PROTECTED] wrote:




 On Jul 8, 4:20 pm, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote:
  This one needs a little clarification.
  Kart racers have been using accelerometer based data acquisition
  systems with trackside beacons for accurate lap times for a while
  now.
 Have you driven a kart? These things GO. This means accelerometers
 have plenty to work with. A person pushing a shopping cart around a
 shop is probably not even raising sensor input levels over noise.


 



-- 
take care,
Muthu Ramadoss.

http://cookingcapsules.com - nourish your droid.
http://mobeegal.in - find stuff closer.

--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups 
Android Discuss group.
To post to this group, send email to android-discuss@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at 
http://groups.google.com/group/android-discuss?hl=en
-~--~~~~--~~--~--~---