A doctoral thesis, submitted to the University of Southampton, that sets out a new way for wireless devices to measure how far apart they are: using ordinary narrow-band radio, no wires and no synchronised clocks.
The problem
Wireless sensor networks (WSNs) are deployed across industrial, scientific and medical settings, built from many small, low-cost, low-power radios. The catch: the readings a sensor produces are often only useful if you also know where that sensor is. On deployment, nodes have no idea of their own position, so the network needs a way to work it out for itself.
Localization breaks into two stages: first ranging (estimating the distance to a set of reference nodes), then computing a position from those ranges. The hard part is ranging cheaply and accurately. At the time, no method could hit the target of 0.3 m resolution and better than ±1.0 m accuracy in line-of-sight conditions using the kind of simple, power-constrained hardware WSNs demand.
The idea
The thesis presents a novel radio-frequency time-of-flight (TOF) ranging system. Rather than chase accuracy with an expensive high-speed clock, it exploits the small frequency difference between the two transceivers involved in a measurement. That difference acts like a vernier scale, letting the system read a sub-clock phase offset and so resolve the flight time far more finely than the raw clock period would allow.
- Two-way ranging removes the need for a shared, synchronised clock between devices.
- Time-dependent, not frequency-dependent: accuracy improves by averaging more samples rather than by running a faster oscillator, which is exactly what low-power hardware needs.
- No extra hardware. The algorithm was prototyped entirely in software on a Texas Instruments CC2431 development kit, using a single channel in the 2.4 GHz ISM band.
The results
The prototype was tested across three very different environments, referenced against GPS and physical measurement:
- Line-of-sight: better than 7.0 m RMS over 250 m, using a 100-sample average.
- Non-line-of-sight: 15.8 m RMS over 120 m through buildings and foliage.
- Indoors: 1.7 m RMS over 8 m, using a 1000-sample average.
Crucially, ranging error stays linear (it does not grow with distance), and the thesis extends the work from point-to-point ranging to full two-dimensional localization of sensor nodes. Performance is ultimately bound by signal-to-noise ratio, signal bandwidth, synchronisation and the frequency difference between devices.
Inside the thesis
The full document works methodically from the fundamentals up: an overview of wireless sensor networks and ranging, the theoretical limits on resolution and accuracy, the design of the prototype ranging system, a prototype locating system, and a detailed analysis of noise, multipath and shadowing effects.
Why it matters to Norse Radar
This thesis is the deep foundation beneath Norse Radar's craft. Measuring distance from the time a radio signal takes to travel (and squeezing precision out of imperfect, low-power hardware through clever timing and signal processing) is the very same discipline that makes ground-penetrating radar work. The research arc runs from an early 2008 prototype through a peer-reviewed journal paper to this doctoral thesis.
