Do you want to track passively the movement of persons in a building without video cameras or movement detectors? Then you should have a look to the Video above. It demonstrates a quite precise RF tracking system set up by the student Joey Wilson from the University of Utah.
In the video Joey explains how the system works on a functional level, rather going into too much technical details. But after a quick search on google I found the original paper, explaining the system in detail. On this link you can find his paper which is called: Through wall motion tracking using variance-based radio tomography networks.
The Joey and his tutor Neal set up peer-to-peer radio network consting of 34 nodes, distributed around the living room of a casual house. Basically the nodes are passing a token in a consecutive order among each other.
- In the first step, the node who has the token measures the signal strength to the other nodes.
- In the second step the node sends the measured signal strength to a central base station where the data is stored and processed on a Laptop.
After the successful transmission of the measured results, the token is passed to the next node.
RF part of a node:
The nodes used by Joey are based on IEEE 802.15.4 (like ZigBee), transmitting on 2,4GHz with a few milliwatts output.
This is how it works:
When a radio wave hits an object, the wave might be reflected or scattered. This is especially the case if there are a lot of obstacles (e.g. chairs, table..etc) and no line-of-sight signal. In this so called multipath propagation scenario, the same signal (because it was scattered) arrives in the receiver several times, but always delayed by a faction of time. The sum of all these multipath signals are added and result in what we call “signal strength”.
Joey makes use of the effect, that whenever there is a change in the scenario (e.g. somebody walking through the room), the signal strength between the nodes is changing as well. Of course, the amount of radio nodes influences the accuracy of the system. In his test he found out that he reaches a tracking accuracy of about 1m with 34 nodes, distributed around a more or less rectangular footage of 100m² (800 square feet).
Once the data is captured, Joey generates a tomography image from the variances of signal strength between the nodes. A tomography image itself doesn’t contain any information regarding the movement of a person. He uses a mathematical filter (Kalman filter) to calculate the movement of the persons. The filter itself helps to simplify the processing. It smooths the effect of noise and prevents the recorded track of movement form unwanted jumping. He explains the theory of the filter quite detailed in his paper. It’s well written and his steps are easy follow if you have an understanding of advanced algebra.
So what’s the use of the system?
It can be used to locate persons or moving objects within a building, without having visual contact to them. The system is working passively, which means that no active tracking devices are necessary. Joey mentions as possible applications the passive location in emergency situations from outside of buildings for police, armed forces or also maybe rescue forces.
I enjoyed reading through his paper. I personally see the sweetness of his work, in the low costs. IEEE 802.15.4 devices are available for very little money. Adding a GPS receiver and increasing the amount of nodes should boost the accuracy in a range between 10 – 50cm.
Maybe I gonna reinstall a copy of Scilab (Open Source alternative to Matlab) and reload the simulations I did a couple of years ago! I feel like being again myself in the phase of writing a diploma theses. I wrote my theses also on a RF, multipath propagation environment