User Characterization through Dynamic Bayesian Networks in Cognitive Radio Wireless Networks

Abstract

Abstract-The current shortage and inefficient use of the frequency spectrum lead researchers to seek technological solutions to this problem [1], thus Cognitive Radio (CR)[2] is proposed, allowing a more efficient management of the existing resources so they can be exploited opportunistically by cognitive users. This paper presents the design and use of a Bayesian network for the characterization of the primary user (PU) in wireless networks (GSM 824.9 MHz) in order to generate a PU activity predictor, which could serve to the central entity of a cognitive network in making spectral decisions. From the results found, it is concluded that the artificial intelligence technique based on Bayesian networks allows to model and predict the behavior of the primary user above 80% for short future lapses of time. Keyword -Cognitive radio, primary user, Bayesian networks, prediction, characterization.I. INTRODUCTION Bayesian networks that are part of artificial intelligence [3], [4] are a graphical representation of nodes interconnected cyclically with each other. The nodes represent a set of variables that forms a system, and which are represented by percentages of probability. The edges represent conditional dependencies between nodes thus being able to have a parent and a son, which concludes in a topology or network structure. This model allows an inference estimating the probability a-posteriori of consulted variables from the a-priori probability of the observed variables, resulting in the classification, prediction and diagnosis of variables of the modeling system. Several research papers [5], [6], [7], [8], [9] reinforce the intention of combining cognitive radio with AI, to apply their methods and techniques in different areas of knowledge and works such as signal processing required for DSA (Dynamic spectrum Access)networks, techniques to improve DSS (Dynamic spectrum sharing)spectrum utilization, spectrum sensing techniques, etc. A CR device interacts with its environment and modifies its functional parameters (auto-reconfiguration) with cognitive ability to meet the quality of service (QoS) required by cognitive users; thus, the CR can make use of artificial intelligence as an intelligent and intuitive form [3] necessary for future estimation of spectral holes, from the behavior of the PUs in the channel.The features of cognitive radio depend on the behavior and activities of licensed users (PU) [10], so that nonlicensed users (SU) should identify the spectrum gaps to seize the opportunities of transmission in order to maximize the use of the band, hence the CR can use it without affecting the transmission of the PU. So the accurate prediction or modeling of PU activities becomes necessary, leading to an efficient use of the spectrum for cognitive users [11] without affecting primary node performance.PU activity is notable for having peaks and troughs determining BUSY and IDLE behaviors and statuses in the burst of transmission; in this regard, statistical and probabilistic predictor models have been presented ...

Modeling and prediction primary nodes in wireless networks of cognitive radio using recurrent neural networks

“…From Figure 2, it is generally observed that the scientific literature bases the representation of the activity of the primary users with methodologies that have an important computational cost such as [13], [14], [15], [16] among others, being unviable in applications of open field when the conservation of energy is important [11]. An alternative, which could solve these shortcomings by increasing efficiency, are models based on self-learning that provide feedback from their own mistakes to enhance future performance, as is the case with RNNs.…”