3d Tracking

[warz]

im just going to throw this out there and maybe guess that you don't have the equipment, or knowledge of creating the equipment that you'll need for this. maybe you could find something that already exists and is similar to what you're trying to do, and modify its hardware / software to your liking? just to see how this kind of stuff works and then go from there.

[Hdx]

I've done work with Sonar/IR before, But no I have never developed something like this before. Which is why i've asked here for ideas. I would like to get my own implementation of it working before I go look at someone elses solution. No I don't have the specific equipment i'd need right now. Except for if I did it using IR. But meh. Its all in the planning/thinking stage right now. Once I get a solid idea backedup with a bit of research i'll get to work on the physical thing.

I'll look more into RF once my boss has left the building. Looks promising sofar except for the refractions.

[Camel]

I just sort of skimmed the article, so sorry if I'm repeating what's already been said, but here's my $0.02:

It depends on whether you want accuracy or going through walls more; if you want to go through walls, you're locked to RF (and relatively high power - so you'd have to use the 2GHz band, which would probably demolish any 802.11 networks in the area), but your accuracy will severely suffer. I also sincerely doubt that you have sufficient background to attempt such a thing yet; I've been in the wireless industry for two years, and I still only understand the basic principles involved.

I don't think radio is plausible for your situation; but one exception would be to re-use an existing TPS solution. Beware though, they (cdmaOne, 802.11n, etc) are all inaccurate, since they are based on a system where light moves really fast, and reference points are really close. The accuracy of GPS is critically dependent on the fact that GPS satellites are far away.

What it really comes down to is how important accuracy is to you. If you just want to tell what room someone's in, you can get away with an 802.11n system -- and you don't even need sensors on peoples bodies; you can "see" a human in the 2GHz band; there are some intrusion detection systems in development that can run on any WRT-based 802.11n router.

If, however, you need to know accurately where in the room a person is, I think you should just use sound; it's probably your only hope if accuracy is desirable. As iago hinted earlier, there's six orders of magnitude difference between the speed of light and sound, so your prorogation delay will also be six orders of magnitude larger than that of an RF-based system, but it also means that your reference points can be six orders of magnitude closer for the same accuracy.

Have you determined how to locate your reference points? If you build a sensor in to each of the transmitters, you can build an ad-hock relative positioning system, where one sensor is simply defined as the origin, and another as the master reference, and all the other points are given with position relative to the positions defined by the first two.

I'm pretty sure you only need 3, even in 3 dimensions, but i'm bad at visualizing 3d so you might be right.

You need not visualize 3d, just understand that each reference point knocks out one dimension (though the reference frame is not fixed). When you get down to 1 dimension, you still need two reference points - and that should be easy to visualize.

If you have an equal number of reference points as dimensions, there can be two points at which the object could be. In the case of 3d, think of a plane defined by the 3 reference points, and mirror the point across that plane.

RF would be the preferred technology for this application (low power, easy to implement, transmitters/receivers are cheap, low interference issues).

Interference in 2GHz is not low; the thing about WiFi is that the AP doesn't really have to know where you are accurately, just where it looks like you are from its perspective.

Even with 4 reference points, you're going to have dead zones, distortion, and interference. However, with enough redundancy, you should be able to almost disregard those problems. You may have to fill every channel in every room.

[warz]

I still think everyones jumping the gun, here. This is the type of project that could possibly require the creation of your own devices, I'd think. I have no experience in this, but it's just what I would assume. You're not going to get very far trying to write up an implementation without knowing what hardware you're using. You won't be able to start writing anything until you know those specifics. In these projects the physical things typically come before the computer manipulation of them.

That's why I suggested possibly finding something you can toy around with before spending the larger amount of money on the parts to build one tailored to your needs. (or having it built for you)

edit: discussing what types of hardware, or "waves" to use, or whatever, though is helpful. I just read Camels post and it sounds knowledgeable.

[Chavo]

Why the hell would you want to use a 2Ghz RF Frequency?  The 80Mhz band range is considerably more appropriate for this project.  The only reason you think it would be too complex for him is because the world of computer/mobile/etc wireless technology is considerably more complex than simple short range wireless communication between simple devices.  Communicating a simple RF signal is as simple as building a digital signal generator (like a simple program on a microcontroller, generating an analog RF signal (transmitter) and using a receiver to convert it back to a digital signal that another microcontroller can understand.  There isn't a protocol that he needs to learn or a library to implement or a set of security steps he has to follow or any other standards other than staying within the right frequency bands that the FCC has allocated for hobbyist projects.

[Camel]

Why the hell would you want to use a 2Ghz RF Frequency?  The 80Mhz band range is considerably more appropriate for this project.  The only reason you think it would be too complex for him is because the world of computer/mobile/etc wireless technology is considerably more complex than simple short range wireless communication between simple devices.  Communicating a simple RF signal is as simple as building a digital signal generator (like a simple program on a microcontroller, generating an analog RF signal (transmitter) and using a receiver to convert it back to a digital signal that another microcontroller can understand.  There isn't a protocol that he needs to learn or a library to implement or a set of security steps he has to follow or any other standards other than staying within the right frequency bands that the FCC has allocated for hobbyist projects.

You would want to use 2GHz because the hardware and software already exists for it, and the price is justified when you consider the work involved to create such a device.

Also, if you try to detect humans at 80MHz, you'd be committing patent claim infringement. Using the 2GHz band does not require taking the patented approach.

[iago]

Also, if you try to detect humans at 80MHz, you'd be committing patent claim infringement. Using the 2GHz band does not require taking the patented approach.

From a quick look at that patent's summary, it's regarding detecting the *presence* of humans using frequency bands that humans absorb, and placing a transmitter/receiver across from each other to see if a human passes beteween them, not detecting the *coordinates* of humans.

[Chavo]

Why the hell would you want to use a 2Ghz RF Frequency?  The 80Mhz band range is considerably more appropriate for this project.  The only reason you think it would be too complex for him is because the world of computer/mobile/etc wireless technology is considerably more complex than simple short range wireless communication between simple devices.  Communicating a simple RF signal is as simple as building a digital signal generator (like a simple program on a microcontroller, generating an analog RF signal (transmitter) and using a receiver to convert it back to a digital signal that another microcontroller can understand.  There isn't a protocol that he needs to learn or a library to implement or a set of security steps he has to follow or any other standards other than staying within the right frequency bands that the FCC has allocated for hobbyist projects.

You would want to use 2GHz because the hardware and software already exists for it, and the price is justified when you consider the work involved to create such a device.

Also, if you try to detect humans at 80MHz, you'd be committing patent claim infringement. Using the 2GHz band does not require taking the patented approach.

The hardware exists for ANY reasonable RF project and the software requirement is negligible.  The 80Mhz band was not meant to be taken literally.

[Camel]

From a quick look at that patent's summary, it's regarding detecting the *presence* of humans using frequency bands that humans absorb, and placing a transmitter/receiver across from each other to see if a human passes beteween them, not detecting the *coordinates* of humans.

It's much more specific than that - the patent covers detecting presence on the line path between a transmitter and a receiver, which is the only way to detect humans in that band. With the 2GHz band, you can "target" specific locations, which is the fundamental principle that 802.11n relies on in order to work at high speed for multiple clients simultaneously. It requires having 3 antennas, though the optimal positioning of the antennas is debatable.

The hardware exists for ANY reasonable RF project and the software requirement is negligible.  The 80Mhz band was not meant to be taken literally.

RF-based systems are not ideal - signals bounce, skip, and get absorbed. This is not like building a voice radio where you just tweak the position of the antenna until you have a reliable transmission; radiolocation is complicated.

This is why I suggested an 802.11n router as a starting point; 90% of the work is already done, and plenty of 802.11n routers out there have open-source, linux-based operating systems. DD-WRT even has a nice built in 3d radiolocator. Modifying it to search for bodies instead of WiFi clients would be much less work than starting from scratch.

[iago]

It's much more specific than that - the patent covers detecting presence on the line path between a transmitter and a receiver, which is the only way to detect humans in that band. With the 2GHz band, you can "target" specific locations, which is the fundamental principle that 802.11n relies on in order to work at high speed for multiple clients simultaneously. It requires having 3 antennas, though the optimal positioning of the antennas is debatable.

Then there you go -- it's not the same as what hdx is doing.

[Chavo]

You (Camel) are trying to make this way more complicated than it needs to be.  Using an RF signal to transmit data is not a complicated task.  Determining how to do so to achieve a goal while keeping in mind the limits of RF (such as signal refraction, which I already pointed out) is not something so difficult to manage that you need to drastically jump your hardware requirements to support WiFi and use a complex software solution that is designed for much more complex purposes.  I know you work in the wireless industry and it's not surprising that you would approach this task from that angle. I can't stress enough to you how unnecessary this is.

He's trying to locate an object in a room that he can place his own receivers in.  He's not trying to reinvent GPS.

edit: Hilariously enough, there's already a patent for almost exactly the approach I'm suggesting:
http://www.freepatentsonline.com/7019663.html

[Camel]

He's trying to locate an object in a room that he can place his own receivers in.  He's not trying to reinvent GPS.

How does that change the game at all? Radiolocation is not as simple as timing how long the signal takes to get there and triangulating.

I never said that it's necessary to make the system as complex as an access point, but I am saying that any old 802.11n router already has this problem in the bag, and that makes it an excellent starting point. Since some of these are open-source, looking at how they are implemented would be a great way to learn how it's done. I don't see why you're taking issue with that.

[Chavo]

Radiolocation is not as simple as timing how long the signal takes to get there and triangulating.

Obviously?  Implementing a way to determine distance using RF or any other wireless technology should be his primary challenge and focus for this project.  IR and WiFi do not have an advantage over RF here.

[/quote]I never said that it's necessary to make the system as complex as an access point, but I am saying that any old 802.11n router already has this problem in the bag, and that makes it an excellent starting point. Since some of these are open-source, looking at how they are implemented would be a great way to learn how it's done. I don't see why you're taking issue with that.[/quote]
A 802.11n router is a complicated piece of technology that is not trivial to integrate with compared with designing an microcontroller/RF transceiver pair circuit.  Examining an open source 802.11n project may be valuable for theory or to help find a solution, but integrating or modifying such a project would not be simple compared to writing a program to manage a A/D converter that receives an RF signal (perhaps utilizing SCI).

[Camel]

Is the system going to be embedded everywhere? If not, how will it communicate, and with what?

[zorm]

My suggestion is to forget trying to do echo location, you're going to need fancy hardware and stuff for that apparently.

It might be much easier to have 4 transmitters at fixed points that transmit continuously, and then put 2 directional antennas on the person of interest. Find a way to move the antennas so that they move to a location of maximum signal and triangulate your position based on that.

Alternatively, if you're assuming that the object won't be suddenly dropped into your room you can use more traditional methods for tracking your position in the room. Much more simple and easy to implement I believe.