Cognitive resource manager framework for optimal resource allocation

Wireless networks are under constant pressure to provide ever higher data rates to increasing numbers of users with greater reliability. At the same time they are becoming more complex and challenging to manage. Great efforts are being done to make the wireless devices and networks adaptive and self-optimizing in order to more efficiently use the resources and deliver good quality services. High spectral efficiency, environmental adaptivity, user-awareness and energy efficiency are highly desired features in the future networks. It has also become important to support these goals at all OSI-layers in a cross-layer manner. Making the wireless systems smarter has been a matter of research under the cognitive radio (CR) paradigm for ten years now. While CR is a very interdisciplinary and wide topic, including dynamic spectrum access and policies, flexible system architectures, learning, context awareness, cooperative networking, etc., most of the contributions so far have been limited to novel spectrum access approaches and spectrum sensing techniques. Mitola’s original vision on context-sensitive smart radios was a precursor, but the current work has been still lacking precise proposal beyond high-level arguments. In this thesis we study the cognitive radios from a system point of view focusing closely on architectures, techniques and algorithms that can enable intelligent operations. We propose a modular cognitive resource manager (CRM) framework, which can facilitate a development of complex control and optimization techniques for resource management in wireless networks on diverse radio environments and problem scenarios. This work contributes towards bringing cognitive radio a step closer to practical implementation by conducting both theoretical and experimental studies of suitable optimization methods and algorithms under the proposed CRM framework. We study in this thesis automatic and adaptive system configuration mechanisms for different resource allocation problems. As most of the problems have heavy optimization phase and often exhibit complex and non-linear parameter dependencies we have studied the use of heuristic algorithms. Genetic algorithm optimizer for PHY and MAC parameter selection has been developed and tested. For autonomous channel allocation we have studied two different classes of algorithms. An approximative coloring algorithm and a corresponding protocol were designed and successfully implemented to minimize the interference in wireless local area networks. An evolutionary game theory method based on balls and bins problem was subsequently developed to jointly address channel allocation and load balancing problems. Finally, the work in this thesis concludes by applying Minority Games to medium access control problem in order to enable self-organization without information exchange overhead.