Fuzzy Logic for Cross-layer Optimization in Cognitive Radio Networks

Abstract

The search for the ultimate architecture for cross- layer optimization in cognitive radio networks is characterized by challenges such as modularity, scalability, complexity and interpretability constraints. In this paper we propose Fuzzy Logic as an effective means of meeting these challenges, as far as both knowledge representation and control implementation are concerned. I. INTRODUCTION Cognitive Radio devices, in their original definition, have the primary objective of providing wireless communication capabilities which are able to adapt to the needs of the user (1). One of the key challenges in this task is to be able to exploit the resources available in various scenarios, using different wireless technologies. Given the state of the art in communication systems, this resource optimization strategy often translates into the need to break the traditional approach of protocol encapsulation, in order to allow for information exchange and interactions between the different layers of the protocol stack. For this reason, cross-layer optimization is nowadays widely accepted as a fundamental component of wireless systems (2), (3), and is expected to play a major role in Cognitive Radio systems as well. In the last decade, cross-layer optimization strategies have been widely studied and adopted, but in most cases the aim was to achieve performance enhancements in specific scenar- ios. Most early cross-layer work dealt with a fixed combination of applications, protocol suites and wireless technologies, and had as its main goal the optimization of multimedia applications or transport protocols over a wireless link such as 802.11 or GPRS. In more recent years, much effort has been put by the research community in trying to synthesize all this experience on cross-layer optimization into a more generalized, universal cross-layer architecture, with the aim of providing all the features and benefits of cross-layer information exchange and interactions to arbitrary combinations of applications, proto- cols and wireless technologies (3)-(7). The ultimate Cognitive Radio is expected to be able to interact with this architecture using optimization algorithms and Artificial Intelligence (AI) techniques in order to exploit the available resources at their maximum, and therefore to enhance the service quality per- ceived by the user. However, although the awareness of the need for a generic and universal cross-layer framework coexisting with the tradi- tional protocol stack is fairly well established in the research community, a definitive formulation for it is still lacking. There are several challenges in the design of a cross-layer architecture: modularity, affordable complexity and scalability of the architecture are of primary importance. In this respect the correct choice of the semantics of cross-layer information and commands to be exchanged becomes crucial: only by rep- resenting information in an abstract, technology-independent format can the modularity and scalability constraints be met. Moreover, if all available information and tunable parameters are exported from all layers to, e.g., a central cognitive engine, the design of the engine itself can easily become impractical because of the overwhelming complexity; as a consequence, it is often wise to have each layer itself handling its own complexity, exporting through the cross-layer system only a small set of highly significant pieces of information. Finally, the way information is represented should allow an easy and unambiguous interpretation, taking into account factors such as data uncertainty and incompleteness. It is our opinion that using incomplete knowledge represen- tation and qualitative reasoning can be an effective strategy in meeting the above mentioned challenges; in particular, we propose the use of Fuzzy Logic as a convenient knowledege representation scheme for cross-layer information, and of Fuzzy Control Theory as a suitable AI technique for the im- plementation of cognitive cross-layer optimization strategies. The rest of this paper is organized as follows. In section II we will review some important issues in Cognitive Cross- layer Architecture design, highlighting the key challenges and showing which ones are not met by existing proposals. In section III we will very briefly summarize the main concepts of Fuzzy Logic and Fuzzy Control Theory, and in section IV we will propose our Fuzzy Cognitive Cross-layer Architecture. In section V we will show a possible implementation which uses fuzzy cross-layering for the enhancement of TCP performance over 802.11. Finally, in section VI the conclusions will be drawn.