Is it possible to build GPS "like" system for high-accuracy local position finding?

I am actually a software guy and only know a little about electrical
engineering in general. (Since this crazy idea has come to mind, I am
teaching myself as much as I can.) I do not yet know what I can
realistically accomplish on my own.

I want to build a scaled down GPS "like" system for an area about the
size of
my house. I would like it to be accurate to about 1 or 2 cm. This has
nothing to do with mapping, surveying, or anything related to actual
global coordinates. I'm only interested in the area in and around my
house.

It would incorporate 4 GPS like transmitters and a number of receiver
units. This would allow me to read the receiver unit's relative
position in area surrounded by these transmitters. I only care about
tracking within my house, and possibly right outside, but not any
further.

Based off my limited knowledge of the real GPS, building my own doesn't
seem to be outside my technological capabilities. (At least with a
decent amount of intense self-training.)

This is based off these facts:
The real satellites use RF waves to send these psuedo-random data
streams and the unit compares its local copy of psuedo-random stream
with the "delayed" satellite streams. (Delayed due to distance RF
waves have to travel.) This delay is used to calculate distance and
eventually exact position of the unit. Using four satalites, it can
triangulate position and calibrate it's own clock to be reasonably
accurate.

My transmitters would have to have very accurate clocks, but I could
also calibrate these as often as needed.

So, is this even possible? Is it too far fetched to even consider
trying? Yes, I expect it to be hard, but would a seasoned EE
profesional be able to pull this off?

Real GPS solutions (DGPS) that have sub-meter (let alone 1-2 cm)
accuracy are
very expensive: $30k - $50k.

One other note, this system would be incorporated into a potentially
lucrative product--if it worked. It would be worth a lot
of trouble to build this.

Even further, does something like this exist allready? or would
someone be interested in building this for me? This is a small step to
my "real" idea, which has nothing to do with surveying, mapping, or any
other global positioning problems.

Thanks in advance and I hope this is the most appropriate group to be
posting in. I have posted to alt.engineering.electrical but fear this
was a poor choice.

Any suggestions/comments welcome!!

 

Does it have to work through the walls of the house? If so what are the
walls made of?

[...]

Your transmitters would be easy to run a cable to, wouldn't they? The
clocks could all be synced to a 10MHz or so signal on the cable. This
would reduce the need to buy expensive atomic clocks down to buying VCXOs.

Does this have to work for moving objects?

How long can we average for before giving an answer?

Do you need accuracy or repeatability? Making something repeat well is a
lot easer than making it absolutely accurate.

 

blargg27 wrote ...

Dependent on the wavelength you are using? And the
wavelength you can use depends on what is legal wherever
you are (?)

You could feed all the transmitting antennas from a single
point and calibrate for the exact antenna cable lengths.
(The satellite folk don't have this advantage! I believe
back in the days of Loran they did it this way (a single
transmitter/exciter and long land-lines to the transmitting
sites.)

Perhaps irrelevant to this specific question, but very tiny
GPS receiver modules and cell-phone modules and similar
cool stuff is available at www.sparkfun.com One of my
favorite places to windowshop and make imaginary
designs.

 

This is based off these facts:

Look up the value of c. Compare this to 1-2 cm. Analyze just HOW
accurate your clock needs to be.

Numerous vendors have fielded or are working on fielding COTS systems
with this capability (but not working off these principles) for
applications such as tracking the position of personnel within a
building.

Bosch is one such, for instance, as is my employer.

 

Just hire a guy to look for it.
There is no technology that will find everything/anything to within an inch
inside a house.

 

My missus complains that I can't find anything she sends me to look for.

Ken

 

Theory of Too Many Items. If you have too many items, you can't find
anything.

 

A couple of $50 GPS modules and some smart programming should see you
get to the 10cm mark (be prepared for some horrible calibration issues
due to signal reflectance inside the rooms). Some sort of RADAR may
work I suppose, but if someone asked me to do 1cm level surveying in a
(say) 20m cube, I would be inclined to do it with stereo
photogrammetry instead. Scanning with a semiconductor laser ($5 plus
cost of stepper motor system) may assist.

Either that or buy a 30cm ruler and a 25m tape measure.

Mike

 

If you are working within a restricted area, some of this stuff might do
what you want:

http://www.polhemus.com/

 

I have thought about this a great deal. In theory, it could work. In
practice, I don't think it would work. The problem is that inside a
typical house, in many cases a non-direct path will have a stronger signal
than the direct path, thus fouling up your distance calculation. In a big
room, like a warehouse, you still have the floor and ceiling to create
multiple paths, and the interference can change the amplitude and phase of
your carrier, which, ultimately, will also throw off your distance
calculation.

In a warehouse, if you could line the walls and ceiling with RF absorber,
and keep the transmitters and receivers near the floor, it could probably
work. But these are inconvenient/expensive/unrealistic restrictions.

If you envision this working in a large open area, you might consider
ultrasonics instead of RF.

Ha ha! I think you are underestimating the sophistication of GPS.

Well, there is something called a GPS pseudolite. You should search using
that term. They are not cheap at all.

See if ultrasonics could work for you. Check out hexamite.com, intersense,
and others.

--Mac

 

You would think she would finally give up, wouldn't you? ;-)

--
Link to my "Computers for disabled Veterans" project website deleted
after threats were telephoned to my church.

Michael A. Terrell
Central Florida

 

What you are looking for is a :

Pseudolite Applications in Positioning and Navigation.

google found lots of hits, the first one looks good.

http://www.gmat.unsw.edu.au/pseudolite/

 

did u consider Sound source Localization techniques?

CMOS

 

1 cm at speed of light is about 30ps. Not impossible, but quite a
challenge. Good Luck. Go ultrasonic, on a per-room basis. Much higher
likelyhood of success.

It's known as GPS pseudolites. As others pointed out, the multipath
situation is going to be really bad. Forget 1cm. If you must do it,
use 4 pseudolites *per room*.

Speaking as someone who did it. Design cost around $15M, a good
sized team, big company, 5 years. I don't want to stop you. :^)

90% of the effort is to get at the last 10% of performance. If
you know everything you need for GPS, you could probably get
your first position fix within a few months. Assuming you buy
a RF frontend and learn first how to do digital designs in FPGAs.

That's correct. At about the same level as "a car uses an engine
to go faster and breaks to go slower" is correct. I'm sorry for
being so condescending.

The main problem is that a simple compare of the PRN sequences
is not possible.

First, the signal is weaker than the noise, by about a factor
of 100. GPS is a CDMA system, you need to know what the delay
is in order to demodulate the signal. This is done by searching
and later tracking the code and carrier phases.

Second, if you want to measure a delay directly, the resolution
depends on your clock speed. For 1cm, that's 30GHz. For GPS,
most receiver do all the signal processing at less than 10MHz
clock speed for power saving. The resolution is achieved by
having fractional PRN generators and integrating over a relatively
long period of time.

As someone else suggested, clock accuracy is not a problem. Simply
run all pseudolites from the same clock via cables.

The problem is clock stability. The exact frequency doesn't matter,
but it has to stay reasonably constant within your integration period.
The more accuracy or sensitivity you want, the better the clock.
Standard GPS accuracy and sensitivity can be done with TCXOs, if you
want much better, the other end of the scale are atomic time standards.

Here is a wrist-sized model. ;^)
http://www.leapsecond.com/pages/atomic-bill/

There are at least 50 independent receiver designs out there.
It's most definitely possible, but you'll have to dedicate a
significant part of your life to do it alone. :^)

Put a price tag on it. For a few megadollars worth of training,
effort and infrastructure, you can design yet another GPS receiver.
Double that, and you may be able to also come up with the pseudolite
design and get everything down to 1cm accuracy, in good conditions.

You can buy indoor positioning from people who have done it already.
It all comes down to how lucrative. Companies design a GPS receiver
from scratch because they hope to sell millions of them.

Google for "indoor positioning". Quite a few products, the cheap
ones seem to use ultrasonics.

Kind regards,

Iwo

 

'splains everything...

[snip]

That makes quite a lot of sense, you don't know squat so therefore you
can do it. Finding your butt would be outside your "technological
capabilities"- go take a hike, file head.

 

Thank you for the responses.

I answered the most relavant questions:

I need repeatability over absolute accuracy. The tracking movements
will be traced as a path and they should appear relatively continuous.

Yes, it must work through walls. The walls can be any lightweight
matierial (wood,sheet rock). The house is really an analagy for the
real area. Their will be no wires or insulation in the real walls.

Can be 2.4GHZ, 900MHZ, and I'm sure others. I'm in the USA.


It needs to be wireless so they can move freely. (I am exploring a
wired system, but won't go that direction until I exhause the wireless
solution first.)

I've done a few rough calculations and a nanosecond clock would
suffice. I am limited to how close I can be to my four
"transmitters"--anything less then 3 meters is too short a distance.

I am aware of clock stability, and don't know how to address this yet.

Well, their is no technology yet! DGPS is accurate to 1 - 2 mm . It
costs $30K - $50K and I cannot afford this. I am still researching and
half the battle is knowing what to look for.

I'll look into ultrasonic and pseudolite, but would this work for
multiple clients roaming around simultaneously? I'll soon find out.

Last question: I'm desperate now...
Is their a publication on GPS technology? Is it an open standard?

-Ben

 

I thought more about this!

How about:

No walls, large wharehouse, all transmitters hardwired together (to
sync clocks).

Direct line of site from transmitter to receiver.

Would it be possible then?

 

Pretend I was a seasoned EE professional. Could I build this on my
own? Without huge budget and lots of expensive equipment!

The one thing I have an excess of is....time!!

 

But be sure to compensate the propogation time over the cables when it
varies due to temperature changes.

Jim

 

If you can live with fairly poor accuracy and can have more transmitters,
your life will be easier. If the receiver has a good signal from 3
transmitters and your transmitters all share the same reference time base,
the phase relationship between carriers gives you a fine resolution
continuous function of position.

Wood and sheetrock act enough different from free space to be a problem to
the system. I see no reason you couldn't add more transmitters if needed
to solve this however.

Ultra sound is fine unless there is a wind.