Slashdot Mirror
MIT Develops Accurate System For Tracking People, Objects Via WiFi (softpedia.com)
An anonymous reader writes: MIT's Computer Science and Artificial Intelligence Laboratory has created a new system called Chronos that can accurately detect the position of electronic devices in a room -- as well as the users who are carrying them -- within tens of centimeters using Wi-Fi signals only. "Chronos works without the aid of any secondary sensors, only using a technology called time-of-flight calculation, which measures the time it takes data to travel from the WiFi access point to the user's device," according to an article on Softpedia, citing a paper (PDF) that the researchers presented at a USENIX symposium in March. "MIT researchers say that by multiplying the time-in-flight value they receive from each user with the speed of light, they were able to detect each user's distance to the central Wi-Fi access point."
Good thing I don't have a WiFi chip implanted into my skull. And I'm not glued to my goddamned phone, so MIT will need to step up its game if it wants to track ME.
slashdot: A failed experiment.
Without bothering to RTFM, does this mean that if they have two (or more) WAPs and the device was connected to each that they could get an accurate location? As it is, it seems that by using in-flight times, they can only determine distance from the WAP which isn't so much a location but a locus of them.
active or passive?
All of the major WiFi equipment vendors (Cisco, Aruba, etc.) have offered this for some time -- though they don't claim anywhere near the MIT Lab's level of accuracy. For instance, Aruba calls their offering "ALE" or Aruba Location Engine. It sits as a separate virtual appliance and communicates to the central WiFi controller (AirWave in their parlance) or to the individual APs if they are operating in autonomous mode. It gets signal strength indications for each WiFi and bluetooth antenna in range of the APs (note: *not* just those devices that are Associated with the WiFi networks served by said APs) and feeds that into ALE. From there, you can map out the devices. Both Cisco and Aruba's products have very extensive APIs to access this info. Maybe they can enhance their offerings with MIT's new technology and get the location resolution improved a bit. For now, in the wild, it's often difficult to get a station (i.e. device) location down to better than a range of 3-10 meters.
Don't think this actually does localisation, just ToF-type distance. So it can tell how far you are from the access point, but not direction (unless they are doing something funky with the multiple antenna). In the main this just seems to be using multiple wavelengths to sharpen up and remove ambiguity from measurements (eg you can tell the phase angle, but need multiple such measurements to know absolute phase difference and thus precise range.
I recall seeing the timing between stations on a wireless network being a common tactic to secure a network. If a station did not reply within a window defined by the time it would take for light to travel plus the time it would take for a secured device on the network to compute a reply then the packet was discarded. I'm sure that there are other methods to respond to such a packet beyond merely discarding it.
They seemed to make a big deal out of being able to do this with a single access point. Reading further I notice that they use other Wi-Fi devices on the network to compute a location which means that with only a single device on the network the ability to determine the location of that single device is diminished. To assure the location of a device they'd still need multiple access points and/or multiple client devices.
I assume that they take advantage of the MIMO capability of Wi-Fi devices that did not exist only a few years ago. This again is much like having multiple Wi-Fi access points, just treat each input and output antenna as a separate device and compute the location that way.
Perhaps I'm missing something important here but I'm not impressed.
I am armed because I am free. I am free because I am armed.
The only reason they get to call this an invention is because they're MIT... Phase detection/time of flight using multiple frequencies is nothing new; main limitation is the shitty clock most things have. Combined with the fact that you need fairly good signal chain components to do it properly. This system will still fuck up I'd guess when large metal objects come in play.
Why do you think there is free wifi in airport?
Is there any better place to spy high profile foreigners' devices than in airports?
So it can detect my casio watch can it?
Misleading article.
From the paper cited, direction is calculated assuming 2 antennas on the access point and 2 antennas on the client ... and then doing an optimization to find the most likely location. When it is correct, you have good accuracy. So the most interesting part is the calculation of time of flight as that is tricky over small distances given how fast light travels.
If they wanted to make it directional without all the caveats, simply beamform the wifi from the access point, and have it scan in bearing, the way radar of sonar does.
"Consensus" in science is _always_ a political construct.
In order to do 10 cm ranging (or "time of flight," as in this paper), you need (with a reasonable SNR, say 10), to have a total bandwidth of order 300 MHz. The 802.15.4a UWB standard provides better than 10 cm ranging, but with a bandwidth of 500 MHz, considerably wider than the bandwidth of a 802.11 (WiFi) channels. especially for 802.11b and its successors at S band (2.4 GHz), which are no more than 40 MHz.
This new Chronos system reproduces some of the technical capabilities of 802.15.4 using, not wider channel bandwidths, but a wider _spanned_ bandwidth, as in Very Long Baseline Interferometery:
To do that and get good ranging from it is a impressive technical feat, as you have to maintain phase coherency across these channels (i.e., you have to know the relative delays imposed by switching between channels, and you need to calibrate these to 0.1 nanoseconds or better).
This may make 802.11 WiFi competitive with 802.15.4 (both UWB and Zigbee); both are communications protocols, but 802.15.4 had, until now, much better ranging performance, which was needed if you want to locate, say, machinery in a factory. Given that everyone understands WiFI and tends to have it installed already, this could really give 802.15.4 vendors problems,
The Xirrus Wi-Fi equipment has this capability built in. With the directional radio's the location has accuracy has always been this good. I don't see why MIT copying features already available to the public in existing technology is news. It is fine that some collegian wrote software to do the same thing as an existing product. But as someone else pointed out, all enterprise level Wi-Fi products have a version of this already.
We did this as a joke about 6 years ago at work using Fritz USB WiFi and Bluetooth Dongles. Nothing New here.
IT Admins Group: Where you decide the content