Position sensing for small, indoor arenas

Tho the subject line states "indoor arenas", has "inside building" now include "behind closed doors" ??

hamilton

Ah, so that's just like any other "watch to see which doorway he went through, last" solution. I.e., it wouldn't "startup" reliably unless you constrained the mobile unit's initial position (i.e., "lose signal for a few seconds")

There would still be the potential for ambiguity -- depending on the geometry involved. Some signals could be identified as bogus/reflections just because their time-of-flights were too long to "fit" in the area being monitored. Others couldn't be dismissed the same way.

Exactly. If you are going to spend money, spend it on components, then labor, finally planning. I.e., if you can make something a wee bit more expensive but eliminate labor (and planning), that's great! If you can spend *labor* to save planning, that's good. Worst case is having to do site surveys and tailor an installation.

Well, gunshots aren't continuous -- "short pulses at regular intervals" -- but I'm pretty sure they would be hard on the ears! :> Just because you can't hear it doesn't mean it wouldn't be damaging (?)

Ah, didn't even consider the glass breaking issue!

See above.

If you have to plan and do a "custom installation", then, chances are, you have to have *your* people do that. Or else prepare to deal with folks who *claim* they "did everything you told me to do (and it still doesn't work)". Either outcome makes the product (and supplier) look bad.

That would depend on the distances involved, no?

I would be more concerned about other "absorbers" in the signal path than the air itself.

Actually, it is close to the 2nd power of the frequency, and strongly depends on humidity, too.

Long while ago I worked with ultrasound in the air. This is only few meters of distance without getting into the ridiculous power levels.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Yes, Don corrected that later. He didn't really mean an arena but more a building. So there will be walls and all that.

Thing is, you don't always know what's inside those walls. In our case RF would perform rather badly because the builders wanted to be extra good and used aluminum-backed insulation literally everywhere. Now we have nice sound-proofing that way but anything RF has a hard time.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Thanks! I'll offer that up and see what comes back

Well, its suitability and application can vary with different environments. E.g., in "open spaces", overhead mounts could be effective at isolating small "location cells" with minimum risk of interference from objects positioned (and moved) within that space. Consider, for example, a commercial environment; floor displays, etc. move around often but the "view from above" is rarely obstructed by these reshufflings.

In a *home* environment, a single transceiver in each "room" could be sufficient to resolve the location to "this room" vs. "that room" (assuming there is some global entity coordinating pings, etc. AND some assumptions on the geometry and scale of the house that can be used to invalidate reflected signals).

The commercial deployment would be relatively easy (dropped ceilings, etc.). The home deployment would be expensive (component costs) but could be as simple as "plug one of these into a wall socket (e.g., PLC/ZigBee for communication infrastructure) in each room".

Note that the capabilities of such a system would vary withthe deployment :-/

Or ZigBee or any other low power beacon...

That's the approach I am taking at home (to get a feel for the issues involved). But, I have much greater flexibility, there, in terms of deployment (more along the line of a commercial establishment) which makes anything learned in this exercise less applicable to a "real" home :-/

Unfortunately, to do it well requires being able to modify the software in the "remote/mobile" unit.

Thanks! I gave it a quick once-over and will dig through it at length -- hopefully this evening.

--don

Exactly. In addition to the differences between the obvious commercial/residential/warehouse environments, construction practices change over the years and in different regions (ignoring international issues for the time being). E.g., this house is stone, block and wood construction -- different parts of the house using different technologies. All of the new framing that I have done is "metal". Some of the drywall is foil backed. As is some of the insulation.

Some homes in the area have *18* inch thick walls. Some homes are built with straw bales.

Some homes have 30 ft ceilings; others a mere *7* feet (consider how sound propagates in a "cavern" vs. a "cave")

etc.

My original thinking was biased towards local beacons -- i.e., "the target is near(est) to beacon 23". This would let you tailor your deployment to meet your needs (i.e., if you don't care about room X, then don't cover it with any beacons!).

It *seemed* that this would be the easiest to deploy -- cost scales with the complexity of the "layout" (number of beacons, etc.).

However, I now wonder if the more "scholarly" approach of overlaying a "location grid" over an entire area isn't the more prudent system? The costs are then (relatively) fixed -- 3 or 4 beacons (assuming they can blanket the entire area) -- though those bits would be more complex and expensive.

Then, train the system to know what you *care* about. I.e., you don't care that the target is 8.3 feet south of the northwest corner of the perimeter but, rather, that it is *in* the region that has been "trained" as "room A" -- regardless of whether room A is 2 inches from room B or 2 *miles*!

[...]

The beacon system can work but we'd have to look at some requirements:

a. Can each person be given some sort of wand? b. What's the rough cost goal? c. At how many installations/year?

Coarse positioning can be feasible since in the US we are allowed 1W at

900MHz. That creates a whopper of an RF carrier. Now one could measure round trip burst returns from the wands, one by one. After all 3-4 beacons have done so a computer would crunch all this and render the locations and code for each.

Learning would be required to improve the precision, to deal with weird walls of other metal objects. Someone would sit at the computer, the other walks around with the wand and a phone, and reports his/her location every 5ft or so. That gets entered and the computer calculates a correction matrix. Of course this falls apart the instant someone moves a bunch of huge steel cabinets or something similar.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

The "remote/mobile" unit (person, device, etc.) has *some* part of the system in their possession.

That depends on the market/application. Targeting a home environment, you have much less $$ to play with. OTOH, a commercial establishment has deeper pockets.

See above. Obviously, less "commercial" than "residential" opportunities. But, the solutions may differ to fit the price points/quantities.

I think of the mobile unit as being passive/receiver. I.e., putting a 1W xmitr on a person gets into health/safety issues...

Ah, but that's where the "coarse" requirement (see previous posts) comes into play!

I don't care about where *exactly* you are located. I don't care if you are "3 ft from where I was a moment ago". etc.

If all I care to do is note which "general area" the mobile is located (e.g., "in the bathroom", in the living room"; "in the automotive department", "in housewares"; etc.) then training can be much simpler.

For example: "I am in region X and will be so until I tell you ("you" being a device!) otherwise". Then, wander around region X while "system" makes "observations" (whatever those entail). Finally, "OK, I am now leaving region X".

Lather, rinse, repeat.

This *should* give the system enough empirical data that it can deduce a return to "region X" any time it sees any signal pattern resembling those observed during the training session (or signals which "fall within" the limits derived from those observations).

If the system ever sees a signal pattern "outside" (whatever that means) the training data set, it can (worst case) *ask* the mobile unit, "Where are you? In housewares or automotive?" and use this observation to augment the training data set.

(see where I am going with this?)

Ok, does that mean you have to live with something they already have or can you prescribe that device?

Eventually you'll have to put real Dollar-numbers behind that/

Understood.

Passive will be very, very difficult over distances of more than a few feet.

Not necessarily. I think cell phones can blast in excess of 500mW and your wand transmitter would only have to transmit at a very low duty cycle. Unless you want to accurately track and clock li'l Joey doing full speed on his plastic tricycle :-)

Ok, that sounds doable but it isn't a small dunk at all. Usually this boils down to time-of-flight triangulation. I have a brief article that I could send you but I'd need your email address (just send me a ping, my reply-to address works). They got a 6ft RMS error which may be too high for you. But then there is bayesian range filtering and all sorts of tricks to improve upon that.

Yes, a learning system which should be low in intrusiveness, meaning self-learing as far as possible.

This could also re-learn moved obstacles. Along the lines of "Wait, no, it's different now but this house only has a straight hallway down there so he could only have gone to the loo :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

[...]

Yes, a retail outlet scenario would surely require that because almost every shelf contains long pieces of metal.

Ok, but then you are way beyond the cost of your idea of 3-4 beacons. In a typical American home it could mean 15 or more. Don't forget the garage and the loo.

If you want to use Zigbee and/or go international with this you'd have to. This is because else your signal strength could become insufficient if there were only 3-4 transmitters.

So you'd have to ping the room motes and those in turn would then ping the wands, measure the time-of-flight and return that info to the command post.

It shouldn't if you are able to get at the RSSI value. And even without it you still have the RSSI of your own RX. If you want to do TOF then you'd need to have a dead-certain number for the answering latency. If it varies then this device is no good to TOF.

Pour a nice Irish coffee first :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Yes, but you would only need beacons in the "rooms of interest"!

Yes, signal strength is relatively easy to get -- even in bogounits, etc.

But, if you want to create a better protocol, it rules out COTS devices. (this is a big drawback so has to come with big *payback*!)

Not fond of coffee :-/

It seems like they are trying to work around existing technology (instead of inventing new). I think this might be worth a more detailed look -- though I am not sure if it would scale nicely. (of course, it can very well be that I need to come up with different solutions for different environments :< )

I have some influence. Ideally, it would be nice to leverage some existing accoutrement (e.g., the BT snooping poses less burden on the user as he may already be "BT accessorized"); barring that, an existing *appliance* (i.e., something that is already "miniaturized"); barring that, something home-grown :>

Sorry, I didn't intend "passive" in the "literal/techno" sense. Perhaps "slave" would be more appropriate.

Get with the times, Joerg -- it's no longer a plastic trike ("Big Wheel"?) but a MOTORIZED SCOOTER! :-/ (sheesh! And to think I had to make do with a "little red wagon"...)

ToF could "economically" work if you move down to AF ranges (hence my initial interest in ultrasound). At RF, I think you need to look into GPS-ish technology :-/

watch your mail

I am *sure* the solution won't be a naive, straight-forward "textbook" approach (e.g., the fixed beacons and surveyed site concept I mocked initially). Rather, the solution will have to exploit things that are overlooked or taken for granted, typically.

Exactly. You can always *ask* the user *if* you are unsure of a *deduction* that you are about to make; but, you shouldn't annoy the user by asking needlessly (i.e., the difference between a well-conceived system and a naive approach to the problem).

Note that there are lots of observations you can make that are easily glossed over -- yet incredibly obvious, once considered. E.g., usage *patterns* could be exploited ("It's 9AM... Bob always goes into the break room for coffee so the current ambiguity in his location is most probably resolved as BREAKROOM and not WOMEN'S LAVATORY").

[...]

The Radio-Flyer wagon? Those are cool. We didn't have those but my grandpa made us wheelbarrows, from scratch. I still have mine, it's at our front porch, with decorative flowers in there. Around 50 years old and looks like new. Ok, the tire has a minor visible crack in it but that's it.

[...]

You aren't restricted to the AF ranges. Many options there. For example, you could do fake data transmissions, change the repetition rate until it fails, and then you'd know the distance. This inadvertently happens with out "modern" DECT phones here. It seems they failed to provide a variable data rate in the standard so there are areas where there's plenty of signal but the link collapses.

Ok, I will :-)

Yup.

This is where TOF can excel. The system could recover after having lost a target for a few minutes.

In some places in San Francisco I would not be so sure about that :-)

But yes, there's lots that can be used. Consecutiveness for example. Certain areas must be crossed in order to get to certain others. So even if Bob goes outside to have a smoke instead of into the break room it won't be plausible if his signal was not picked up in the large hallway.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

2010-10-25 23:29, D Yuniskis skrev: 802.15.4 Radio + measuring of RSSI ?

Best Regards Ulf Samuelsson

And "Flexible Flyer" sleds (always felt sorry for the folks who got stuck with toboggans -- though in an ice storm it was a toss-up as to which was more of a thrill!) and wooden puzzles...

(sigh) Gee, I wonder how we ever survived without iPods, iPhones, iPads, The Internet, etc. :-/

Solid tires instead of these soft, spongy inflatable ones?

Ah, I hadn't thought of that!

IMO, you have to be *really* unobtrusive if you want to gain any "acceptance". E.g., if the "mobile" unit kept prompting the user: "where are we? I can't sense my beacons!" then the user would quickly learn to *leave* the mobile unit "someplace convenient" and *fetch* it whenever he/she needed to interact with it -- thus defeating the purpose.

Understood. You have smarts available. You can store *lots* of state. *Learn* from that and make smart deductions instead of just looking at the "here and now"...

Reminds me of this scene:

I still do.

Well, ok, that I need. But only for the job.

Yep, and white-walls no less. But I'd have to shrink back to 2-3ft tall in order to operate this wheelbarrow.

Essentially it would work like the vernier scheme on old style calipers. Before they came out with the dials on them, and then the new highfalutin digital ones. Still haven't made that step yet.

[...]

I can imagine that. Major Tom to base: "I'm in the loo, darn it. Why on earth do you need to know?"

I am not sure where this is going to be implemented but don't overlook the mischief potential with some folks. "Hey, if we leave this thing in the entryway, string a rope, walk to the meeting room and then yank it over really fast we could fry the computers" or "I'll stand in the door way over there and you toss it to me from the railing over yonder".

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

I write *really* robust code EXPECTING the unexpected. E.g., if I am driving a motor "left", I also watch for the "right" limit switch which, in theory, "/* CAN'T HAPPEN */" -- unless the cable to the motor has been attached incorrectly!

I have a penchant for coming up with odd circumstances that, when ignored, cause code to break (you *really* don't want to tell me your product is "100% debugged" as, historically, I can find a fatal flaw in less than 15 minutes: "How did you get it to *do* that????!"). Of course, turning this "ability" back onto myself, I anticipate how *my* devices can break -- while I am still in the design phase. (bug fixes are very expensive :< )

Same here. At my last employer I was "Mr.Crash-it". Sometimes the SW guys came into my office and asked if I could take a new or upgraded system for a spin. Sometimes half an hour later 3-4 SW engineers would stand around a solidly frozen machine, scratching their head. "This was not supposed to happen!"

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.