I know that bluetooth positioning has been done many times. But I’m not exactly sure on the best way to set it up.
I want to setup a bluetooth positioning network on the whole floor of a building. There are various walls and obstacles that hinders the bluetooth signals, but that wouldn’t be the problem if you have bluetooth antennas with extremely high power.
I want to go simple and stick with having only 3 antennas and a central computer receiver that takes all the data from the antennas and processes them. The antennas might run off of pics or stamps and probably will have to be longer ranged than the smirfs to have strong enough signal.
Does anyone know a good bluetooth antenna for this purpose? It will have to be able to connect to some kind of board, anyone have a recommendation for that too?
I’ve seen some people do that bluetooth triangulation with signal strength, but for my purpose I do not want to depend on the signal strengths, especially in an environment where they can vary significantly. In other words, is there any possible way to figure out the distance between bluetooth devices without using the signal strengths? It doesn’t matter how many extra antennas that it would take.
I’ve used products that do this for WiFi, not Bluetooth. The tracked device can be a special WiFi tag, or most any WiFi device.
The techniques to do this are well known but a bit to complicated to explain here. A leading product that does this “RF Footprinting” is from Ekahau. Another, Aeroscout, does the same, and optionally, TDOA.
In TDOA, three or more costly receivers measure with nanosecond precision the time of arrival of a pulse from the tracked device. Kind of like triangulated radar. Also called multilateration.
$$$$
The RSSI approach is what’s commonly used. Relies on heavily overlapping coverage of 3+ access points and a survey data set that gives a priori information on the effect of walls, large furniture.
A big shortcoming of all these is that the Z axis is quite ambiguous because ceiling heights are low (like 4m) in multistory buildings, compared to the measurment accuracy of, say, 5-10m.
So estimating the correct floor number takes hueristics, not simple location.
Yeah I wasn’t very partial to TDOA either. It requires quite alot of money to achieve. And RSSI doesn’t seem to be the way I need. I do not want to take into account the various obstacles in my data.
Both methods above solves for the distance between an antenna and a device. What if I just use a sparkfun GPS or something to find distance? I’ve never worked with the GPS before so I’m not exactly sure how accurate they could be. Would this be a cheaper and better alternative?
EDIT: I’ve looked into GPS a bit more and I’m still not sure about the best components to use. I have two approaches that I could use:
Every bluetooth antenna also has a GPS antenna and the central pc is the only gps receiver
Every bluetooth antenna has an antenna and a receiver, so they could just directly find their relative distances from each other.
Method 1 assigns the task of calculating position to the central pc while method 2 demands a degree of processing power for each antenna. Method 2 probably will be much more efficient but would take more time and money to construct. Which one would provide the more accurate location?
This solution only allows for the prediction between the antennas not between an antenna and a device, although if the device was fitted with one it would also work. But suppose that the device being tracked is a cellphone- theres no way I can get a GPS working with that. If I know the distances of the antennas from each other, is it possible to find the receiver’s location?
Basically would it be possible to use the GPS’s time to do TDOA?
The GPS’s PPS signal is accurate to within a few nanoseconds (I’ve seen projects where a PLL was used to generate extremely accurate 10MHz clocks using just the once-per-second GPS time pulse). However, you’re still going to need a nanosecond-accurate clock, like a binary counter clocked in the GHz, to actually measure the time taken for a signal to go between the devices.
Furthermore, if you’re using GPS on both devices, why not just use the GPS location to tell where they are? To get GPS time you need a GPS signal, and if you have the signal then you have your location.
Indoor tracking is a complicated field, and there are very few commercial solutions out there. Many rely on ultrasound/RF distance measurements, I’ve never seen one that relies on TDOA alone.
its not a problem to use GPS to figure out the location of the antennas. But I have to know the location of the bluetooth device, which does not come with GPS.
If the minimod is around 1575.42 Mhz. So I was thinking of using GPS time to calculate the time lapse between when the bluetooth signal initiates from the antenna to when the signal returns after detecting a bluetooth device.
If the bluetooth device is 1m away from the antenna, it would take 6.66e-9 seconds for the entire trip. Meanwhile the GPS is updating itself 1.575e10 times per second. At that rate and at that time, it updates 1.575e10*6.66e-9=105 times.
So even from a meter away, the GPS would still be fairly accurate in the time lapse.
You talked about once-per-second GPS time pulse, but isn’t the frequency of a GPS the number of pulses per second? 1575.Mhz would be a lot of pulses wouldn’t it?
Sorry to say so, but it’s not nearly as simple as you think. GPS does not update itself at 1.5GHz. Go spend a few hours reading about the details of GPS and you’ll realise what an incredibly difficult problem this is. Use ultrasound, or if you must use radio then use signal footprinting.
Ah, I see what you mean. It does make more sense that GPS updates only around 1Hz.
I’m not sure I can go with the ultrasound however, since obstacles could mess up the timing, especially in dynamic environments. As for radio, I don’t know if it’d be possible to be used to track a cellphone-is it?
stevec talked about using multilateration, which basically uses more than one receiver. Apparently the more receiver used the more accurate the timing is. I’m fine with having more receivers, but are there any way of keeping nanosecond precision timing without GPS? Ideally I’d like something that I could just attach with some development board and microcontroller. I have a feeling that there aren’t any…
Alternative I can just go down the signal strength path…maybe if I have numerous bluetooth antennas that checks for the signal strength from the cellphone and since I know the positions of the antennas I could somehow come up with a fairly accurate average of the cellphone’s location. Is this a possibility?
Multipath does affect ultrasound, but it affects radio too, you will have to deal with it either way if you’re going with any time-based approach. Ultrasound is so much easier to deal with than radio though, because 330m/s leads to much more friendly timescales than 300000000m/s.
RSSI footprinting is clever, and sidesteps a lot of the multipath and interference problems, in fact to some extent using them to its advantage. You pretty much have a bunch of receivers distributed all around the environment, and you calibrate the system by taking readings with the target at many known positions, then work backwards from that prerecorded data. It’s not easy, but this is a difficult problem.
RSSI footprinting in my tests of best-of-breed professional products, combined with careful database construction (survey) - can yield about 10m accuracy on X and Y - but in a building with lots of walls. Without walls, it can’t be accurate due to the inverse square law of RF propagation.
In a multistory building, the nemesis is the Y axis is ambiguous as to which floor number to call. The best products don’t use simple XYZ solutions; they use heuristics and a CAD drawing of the building. These and a software worm-trail allow one to weight the solution estimate, esp. on Y based on history - elevators, stairwells, etc.
That’s why they get some serious $$ for their server software (Aeroscout, Wherenet, et al).
A hospital doesn’t want the system to say that the IV pump is on either the 3rd or 4th floor, in either room 410 or 310.
But if your expectations are more modest (i.e., no Y to deal with), try using 802.15.4 modules and the new RTLS standards. Cheaper than WiFi and battery life is better. Some '15.4 products support RTLS. Google about RTLS it is an ISO standard
stevech:
RSSI footprinting in my tests of best-of-breed professional products, combined with careful database construction (survey) - can yield about 10m accuracy on X and Y - but in a building with lots of walls. Without walls, it can’t be accurate due to the inverse square law of RF propagation.
About that-I’ve read an article of bluetooth tracking in a zoo, of bluetooth tags using multiple access points:
They are using some kind of custom made access control system for this. I have no idea how it works…are they using RSSI footprinting?
I want to just achieve what they have with only the basic needs. If they can do it, then it must be possible right?
But if your expectations are more modest (i.e., no Y to deal with), try using 802.15.4 modules and the new RTLS standards. Cheaper than WiFi and battery life is better. Some '15.4 products support RTLS. Google about RTLS it is an ISO standard
Using Zigbee seems like a good alternative, but wouldn’t that require the devices being tracked to have certain components added? Whereas bluetooth can just track any bluetooth devices.
With, say, 802.15.4 (ZigBee is not a synonym) - the radios can be cheap enough to put LOTS of them in the infrastructure. This becomes an electronic sign post technique rather than RF footprinting.
Many amusement parks are using Aeroscout’s system. Too expensive for hobbyists.