Previous Abstract Return to Session C1 Next Abstract

Session C1: Sensor Aiding and Augmenting

Chained Wireless Synchronization Algorithm for UWB-TDOA Positioning
Vaclav Navratil, Josef Krska, Frantisek Vejrazka, Czech Technical University in Prague, Czech Republic; Vaclav Korecek, RCD Radiokomunikace, Czech Republic
Location: Windjammer

The UWB (Ultra-Wide Band) is a very popular technology allowing localization of users (tags) in a limited area. Two methods are used: TWR (Two-Way Ranging) and TDoA (Time Difference of Arrival). The TWR requires interrogation of the user equipment with several fixed nodes of the UWB network. This approach requires a lot of air-time. Moreover, considerable energy is consumed by the receiver on the user equipment side.
On the contrary, the TDoA method can estimate user position on the basis of a single “beep” from the user equipment. No UWB receiver has to be active in the user equipment and only a fraction of air-time is required per session in comparison with TWR method. Therefore TDoA offers substantially higher positioning session frequency which allows positioning of more users and higher update rate of user positions. As there is no need for power-demanding signal reception the battery of the user equipment will last significantly longer. However, in order to obtain meaningful TDoA measurements, the fixed nodes of the UWB network have to be synchronized rather precisely, sub-nanosecond accuracy is desirable. This is the critical point of the UWB-TDoA system, since the performance of positioning is inherently dependent on the ability to measure time delays not only precisely, but accurately in the first place.
It is favorable to perform this synchronization in a wireless manner, since deployment of synchronization cables is rather inconvenient and costly. There is no doubt that the wireless method requires clear line-of-sight between the synchronized nodes to provide reliable results. In this paper, Kalman filter-based algorithm for tracking clock phase offset, frequency offset and frequency drift is described. This approach extends the most common solution, where only phase and frequency offsets of the clock are estimated. Analysis of real data will be provided and compared to a filter that omits frequency drift estimation.
For example, frequency drifts are often caused by temperature changes. Rapid changes of ambient temperature and device warm-up represent the most challenging phases. Software estimation and compensation of the frequency drift enables reliable operation shortly after device turn-on, or near the building entrances, where warm and cold air masses are mixed. Naturally, the availability of position estimate is higher. The preliminary results indicate that the available precision is typically below 250 picoseconds, i.e. 75-milimeter range error.
The proposed approach relaxes the requirements on oscillator stability and provides performance margin for oscillator aging effects. Thus, the variety of suitable components is wider and therefore designer can save on dimensions and cost of the device.
In vast UWB networks the direct line-of sight to the node with master clock is not available to all nodes – it is available only for a few nodes in its vicinity. Nevertheless, it is possible to design the network in a way that each node has a clear line-of-sight with some others. Then it is possible to disseminate the accurate master clock information via a chain of node-to-node connections. A method for chaining the algorithm for clock model estimation will be proposed. Clearly, the chaining will introduce negative impact on the achievable timing precision and accuracy. Moreover, equipment delays will have more significant role in this case. Experimental results will be provided and methods for dealing with the precision loss will be discussed.



Previous Abstract Return to Session C1 Next Abstract