Agent based modeling of the spectrum distribution in the cognitive radio networks

Cognitive radio networks are complex adaptive systems, in which the conditions rapidly fluctuate over the time being. PUs and SUs are mutually linked and the interactions show significant nonlinear character (e.g. regulation of the transmitted power). This fact is in strong coincidence with the existing literature, where commonly the standardized mathematical strategies are being formulated (e.g. game theory). The game theory from its fundamental principle lacks the capability of capturing the dynamic evolution of the system (number of SUs fluctuates in the time, regulation of the power, SNR fluctuation) and thus, its application in the real environment is at least, questionable. Based on this fact we can claim that yet introduced freuqency sharing models investigate simplified scenarios (e.g. few PUs and SUs). This could be huge limitation as the recent estimates predict scenarios where ubiquitous wireless connectivity is given, mobile devices are the means for Internet access and seven trillion wireless devices serve seven billion people by 2017. In order to provide a more general modelling solution, we adopt the agent-based modelling and simulation (ABMS) approach able to capture the behaviour of dozens of entities operating in the network. First strength of ABMS consists of its ability to encompass there ports and attributes of individual autonomous entities. They can be easily implemented and quite flexibly changed responding to immediate needs. The second strength involves the emergent behaviour. According to some studies "By modeling large number of interacting agents - each having the own objectives and characteristics and the ability to make decisions- complex macro-scale aggregate dynamics emerge". Observing the emergent behavior, new insights can arise from the results of simulations. The third strength is the relative simplicity and robustness of ABMS. ABMS reduces the complexity of the system control by replacing the aggregated outcome of a phenomenon, which is difficult to explain or to predict, mostly without applying complicated decision rules of self-aware entities at a lower level.

All the above attributes help the natural opening of the application of the ABMS into cognitive radio management. Promising way to design the robust spectrum sharing/trading models in the cognitive radio network is to encompass the interdisciplinary activities. AMBS allows to introduce the new concepts and ideas previously introduced in other scientific branches, such as biology, computer science or economy. From already conducted analysis we believe that there exist several strategies, e.g. Demon and Potts algorithm, or Bak-Sneppen model that could potentially found its application in the process of modeling of the spectrum distribution in the cognitive radio entworks. The models presented within the project frame will be primarily designed to a situation, in which a large number of PUs and SUs are interacting. Since the PUs and SUs are frequently exposed to different types of the operational perturbations (e. .g fluctuation of the receiver power,), a sufficient level of the robustness is necessary to ensure their dynamic stability. In this regard, the usage of e.g. the Bak-Sneppen model can be seen as a blueprint of our thinking framework, partly justied by the fact that evolutionary process are remarkably robust against to a wide class of perturbations. The goal of the submitted project is the introduction of the novel methodologies based on AMBS, more specifically in the spectrum sharing and trading models. We will use the synergic impact of the cognitive radio network structure and AMBS using different algorithms showing huge interdisciplinarity. Primarily, we will focus our attention on the application of Bak-Sneppen model of coevolution, as the evolutionary process are remarkably robust against to a wide class of perturbations occuring in the cognitive radio network. In addition, we aim to take into account the specific behaviour of other algorithms, including Demon algorithm yet used in the termodynamical structures or alternatively Potts model describing originally the molecular structure.