Evaluation of a Belief-based Decision Making in a Real-time Platform for Cognitive Radio Networks

Abstract

This paper presents a real-time testbed based on reconfigurable devices able to transmit and receive data at different operating frequencies, which are dynamically configured. The main objective of this paper is to provide an advanced and realistic framework where the performance of algorithms and policies for spectrum selection in Cognitive Radio (CR) networks can be fully evaluated in a real-time environment. In particular, this paper focuses on a CR belief-based decision making framework particularized for different observation strategies that decide when measurements are carried out to predict the radio environment dynamism. Results show that an accurate selection of an observation strategy is a key element for the trade-off between achieved performance and measurement requirements.

Implementation of Cognitive Radio Networks to evaluate spectrum management strategies in real-time

“…This domain separation enables the development of generic cognitive management functionalities independently from the decision-making process. Knowledge management functionalities that exploit real-time information achieved through awareness processes and stored in apposite databases and that support the decision-making can find applicability in a wide set of scenarios and use cases as demonstrated in the referred papers [20][21][22][23]. Moreover, they provide satisfactory performance results in terms of QoS, QoE and reduction of the signalling, as will be widely explained throughout the next sections.…”

“…In detail, this solution provides a framework to assess innovative spectrum management strategies exploiting the CR paradigm that have applicability in a wide set of scenarios and use cases. The functionalities of this architecture have already allowed achieving satisfactory results as published in referenced papers [20][21][22][23]; notwithstanding, these publications either provide a general overview or address particular testbed issues and capabilities, and they do not provide an overall or exhaustive vision of the presented platform. Then, the aim of this work is twofold: (i) to provide a very detailed description of the whole developed platform, its entire potential and its applicability, with a holistic vision and discussion of the testbed, and demonstrate how it can be useful in a wide range of studies and scenarios; and (ii) to illustrate a particular application in a realistic Digital Home (DH) scenario to assess the practicability of the platform in terms of user-perceived end-to-end QoE, also in a realistic environment.…”

“…In this context, the authors in [22,23,45] proposed a spectrum selection decision-making framework based on the so-called belief vector to predict the environment dynamics, which allows avoiding measurements when possible. In detail, the belief vector assesses the probability that the radio environment is under specific conditions at a certain time based on past measurements (observations).…”

“…Supposing the use of the jth link, r (i) j,k denotes the reward that this link receives when using its allocated SB i and the interference state is S (i) (t) = k. The reward is a metric between 0 and 1 capturing how suitable the ith SB is for the jth radio link, depending on the bit rate that can be achieved in this SB with respect to the bit rate required by the application. The spectrum selection policy proposed in [22,23] and [45], executed at time t for the jth radio link, targets the maximization of the expected reward that the session will experience along its duration D j .…”

“…Moreover, because the observation stage can be very costly in terms of practical requirements, an observation strategy should guarantee an appropriate trade-off between performance and the number of measurements achieved through signalling procedures. Hence, papers [22,23,45] consider the following three observation strategies: tual observations are performed. In this case, it can be easily proved making use of the properties of the ergodic discrete time Markov processes that the values of the belief vector will be equal to the steady-state probabilities [46].…”