Towards self-organizing wireless networks

The concept of self-organizing networks is a promising approach to address fundamental challenges in current and future wireless networks. Not least the dense and heterogeneous nature of the networks, the scarcity of resources, and high costs of manual configurations necessitate efficient self-organization techniques that autonomously adapt crucial system parameters to changing traffic and network conditions. This thesis is concerned with key aspects of self-organizing networks, in particular the generation of the required knowledge, and the design of particular self-optimization mechanisms in wireless networks. Any self-organization functionality depends on knowledge about the current system state. The first part of this thesis deals with learning techniques that enable a wide range of self-optimization and self-configuration functions. In particular, we investigate the generation of knowledge in form of geographical radio maps. The radio maps are generated based on user measurements, which arrive sequentially and continuously over time. To process this persistent data stream, low-complexity online estimation and learning techniques are needed. We employ powerful kernel-based adaptive filtering techniques, which are robust to measurement errors. To demonstrate how additional context information can be taken into account, we show, how knowledge about anticipated user routes can be incorporated in the learning process. Moreover, we investigate the performance of the algorithms in different scenarios; more specifically, we apply these techniques to the problem of path-loss estimation and to the problem of estimating interference maps. Such radio maps are considered invaluable for network planning and optimization tasks. Furthermore, we demonstrate how interference maps can be used to support network-assisted device-to-device communications. In addition to the aspect of knowledge generation, the second and third part of the thesis deal with particular self-optimization techniques. The second part of the thesis is concerned with self-optimization in interference-limited networks, motivated by the trend towards densely deployed heterogeneous cellular networks, which are expected to become prevalent in the next generations of wireless communication systems. Especially in multi-antenna networks, inter-cell interference coordination poses a major challenge, since not only temporal and spectral resources, but also the spatial dimension has to be taken into account. To address this challenge, we propose distributed coordination algorithms for inter-cell (and intra-cell) interference coordination in SDMA-based cellular networks. Based on a local maximization of a network-wide utility function over average user rates, the proposed algorithms autonomously adapt the transmit power budgets of particular resources. Using system-level simulations, we show that, especially in networks with high user mobility, the control of average power budgets for particular time-frequency-space resources is superior to a direct control of the transmit powers. In the last part of the thesis, we consider more general network topologies with a stochastic traffic modeling. We derive a framework to design network-layer control policies based on a suitable cost function. Thereby, the framework adapts scheduling and routing decisions to the requirements of different services and applications, while ensuring queueing theoretic stability. As particular applications, we investigate the cost function based control approach for networks with minimum buffer constraints (e.g., in case of multimedia streaming), and for networks with energy limited nodes. In addition, we show how existing cross-layer control algorithms can be adapted to our control framework.

Author dc:creator

• Kasparick, Martin

Licence dc:rights.uri

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language.iso

en

Identifier URI

http://dx.doi.org/10.14279/depositonce-5062

OAI identifier oai:identifier

oai:depositonce.tu-berlin.de:11303/5387