Diploma Thesis

Neural Network based UWB Geo-Regioning

Ultra-Wideband (UWB) Geo-Regioning

This positioning technique follows a new approach exploiting the Ultra-Wideband (UWB) nature of the channel to achieve a rough localization of UWB transceivers in rich multipath environment. We suppose that the channel impulse response (CIR) of a transmitter/receiver (TX/RX) pair is almost unique, given by the many resolvable multipath components that result from the individual geographical constellation/position of RX and TX. At a certain RX the CIR received from any TX is like a signature of the TX position. If two TXs have a very similar signature they are very close together with a high probability. Although it has been shown, that the spatial correlation of the signatures keenly decreases within about 10 cm, we could show that there remains enough information to decide, whether two signatures belong to the same geographical area/region or not. We refer to this approach as "geo-regioning". We assume that a region can have a size of several dm3 up to several m3. In data aided geo-regioning the positions of some specific reference nodes in the network are known. This information is used to derive from the regioning process the position location information of all received signals. This facilitates a variety of location aware services and protocols in dense ad-hoc networks. Blind geo-regioning does not use a priori location information. It has many interesting applications such as (i) data fusion in dense sensor networks and (ii) routing in hierarchical sensor networks, where the clusterheads may base the routing selection on the region information.

Neural Networks

Neural networks are composed of simple elements operating in parallel. These elements are inspired by biological nervous systems. As in nature, the network function is determined largely by the connections between elements. You can train a neural network to perform a particular function by adjusting the values of the connections (weights) between elements. Commonly neural networks are adjusted, or trained, so that a particular input leads to a specific target output. The network is adjusted, based on a comparison of the output and the target, until the network output matches the target. Typically many such input/target pairs are needed to train a network.

Work Description

Power Delay Profile Power Delay Profile

The figures show measured power delay profiles of channel impulse responses originating from two different regions.

Subject area Ultra Wideband, Positioning, Neural Networks
Type of work 40% Theory, 30% Simulation, 30% Software
Student Luciano Leins
Supervisor Dr. Christoph Steiner
Professor Prof. Dr. Armin Wittneben