Can I locate the object by reading its signal with RFID readers?

Question

Say, I have a moving object [X] inside some territory and a set of long-range RFID readers [<]:

-----[<]---------------------[<]------
|                                    |
|                                    |
|                                    |
|                                    |
|                   [X]              |
|                                    |
|                                    |
|                                    |
|                                    |
|                                    |
|                                    |
-----[<]---------------------[<]------

The idea is to make possible to track cattle or chicken walking around a fixed area range.

Is it possible to get the time signal goes from the object to readers and thus locate it in real time?

How can it be performed?

Answer 1

I worked on a system that aimed to do just this using 900 MHz semi-passive RFID tags (battery powered tag silicon but purely passive communication, it gives you a lot more range since the tag doesn't have to harvest enough energy to power up).

ToF or TDoA (Time of Flight or Time Difference of Arrival) don't work well for RFID. You can do a measurement of phase difference between transmitted and received signal and that will give you the fractional part of the wavelength in your range to the tag but solving the number of complete wavelengths isn't possible with standard hardware (at least we couldn't find a solution).

Our solution was to take a series of readings along one side of the field (it took about 10). The readings weren't all taken from the same height above the ground. We then took the RSSI numbers from the readings and from that calculated the tag location and height. We were taking readings sequentially from the same receiver that was moving but there is no reason why you couldn't use a set of fixed readers.

The calculations ended up being non-trivial. You have the standard inverse square law for the signal strength but you also need to factor in the ground bounce, that will depend on antenna heights, frequency and ground material. You get both direct and ground path signals in both directions giving a total of 4 different signal paths that are interfering with each other. You then also have to factor in tag and receiver gain patterns.

However once you have a good enough model you can create a probability map of where in the field the tag is likely to be based on the observed signals. It was taking matlab about 20 minutes to crunch the numbers on a 10 year old PC. I managed to get the end results to normally be within 20-30 cm of truth at ranges of up to 40 m.

When it worked we managed to get this down to around 5 cm error by factoring in the partial phase information of the returned signal but that was unreliable at times.

The killer problem ended up being that as soon as you added any RF absorbing/reflecting objects into the environment (e.g. people, animals or anything made of metal) then the whole system fell apart.

For the sort of problem you are describing I'd normally say to go with a UWB based TDoA based system. That will give you 10-50 cm accuracy depending on how well the system is dialed in with low latency and be fairly immune to environmental changes if base station antennas are mounted high enough. While not as cheap as an RFID tag you can make the tags in the $25 range and run them off a watch battery of a year or more.

Answer 2

There are many solutions that do exactly what you want. Commercial location beacons exist, but as shown here and here tend to be the newer 900 MHz or 2.6GHz tags.

There is the possibility of making LF and HF tags work by using multiple directional antennas (you could work with directional TX and Omni RX for example), but none of this is exactly easy DIY project work. I'd suggest some reading might help you decide if it's within your capabilities.

You might try 1 a DIY project, or 2 some research, as starting points or search on RFID and RTLS (real time location services).

The complexity will depend almost directly on the resolution you need, so you need to start by working out the specifications you want to accomplish. Others may then be able to help you more.

Answer 3

The bit rate on commercially available RFID readers probably won't give you the timing resolution you need. The propagatin speed of light is about 1000ft/us, so even at 1Mbps your positional accuracy would be ±1000ft.

Instead of using RFID, just have the moving unit transmit a sinewave at some frequency. Then receive the signal from at least three antennas. Take the received sinewaves and compare their phase to find the difference in propagation time. Those differences in propagation time will allow you find position.

An easy way to compare the phases is to run the received sine waves (after the inital RF amplifiers) through a comparator to turn them into square waves. Then run the square waves from each set of antenna through a high speed AND gate. The AND gate will give you the overlap of the two waves. The duty cycle will be 50% for 0° phase difference and 0% for 180° phase differnce.

You can then run the output of the AND gate through a low pass filter with a long time constant (say 1s) to get a DC reading that varies linearly with the phase difference. Read the DC signal with a high resolution ADC to get a very accurate position measurement. Note that the ADC need only read a DC signal so it does not need to be high speed or expensive.

You will need three AND gates, each comparing the difference for a different combination of antennas. All three gates should be in the same package and should have low channel to channel propagation delay differences and skew.

You may substitute the AND gate phase detector with other logic gate types such as XOR if you wish.