A Predictive Model for Cognitive Radio

Abstract

Public reporting burden tor thiS collection 0o intormation is estimated to average 1 hour per response, including the time 0or reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate applications a radio platform not only allows us to tackle the probthat were traditionally implemented in custom silicon to lem of spectrum utilization, but also serves as a terrific general purpose processors (GPP). In the area of wireless platform from which one can maximize the performance communication, this transition has given birth to the field of the system by intelligently manipulating communication of software-defined and cognitive radio. These-smart parameters. CR systems are ideally suited for research and radios, or Cognitive/Sofware-defined Radios (C/SDR), experimentation with spatially aware applications, adaptive can potentially make more efficient use of the available routing, cognitive media access control (MAC) layers, and RF spectrum and adapt to a wide range of protocols mutable physical layers. and environments. One of the key benefits of having aIn our pursuit of developing a viable CR platform, C/SDR is it's ability to change communication parameters we contend that it is crucial to understand how varying in response to changes in application needs and the parameters at the physical, data link and network layers can radio frequency (RF) landscape. While understanding the affect the performance and reliability of a wireless system. effects of changing communication parameters is a critical Understanding how these parameters affect performance precursor to the development of a predictive model, it is allows the researcher to consider how a system might be not the focus of this paper This research builds upon our optimized [1]. We see this as a first step in the development investigation of the affects of varying these communication of an algorithm for cognitive radios. Once the performance parameters through the development of a predictive model. implications of varying these parameters is understood and This model allows a C/SDR to dynamically modify its a model has been developed, one must consider secondary configuration in order to improve system performance.issues. These issues include decisions about when and how to change configurations, how these changes are propagated "The views expressed in this article are those of the throughout CR network, and how much time can be spent author and do not reflect the official policy or position of computing a change in configuration.

Optimizing for sparse training of Cognitive Radio networks

“…For this evaluation, we adopted the setup described in [7] and placed a CR in the scenario shown in Figure 3: the Cognitive Radio in the middle is sending a large packet data stream (1400 byte packets at 384 kbps) and a VoIP stream (120 byte packets at 96 kbps) to the server on the left. The jammer on the right is transmitting continuously at the same frequency with constant power.…”

“…In the analysis of the experiments run by the DOE approach, Weingart et al [7] report that only 4 main factors and five two-factor interactions were statistically significant in an ANOVA analysis to explain average throughput of a CR. In an additional regression model to explain throughput, a selection of main factors and two-factor interaction achieved near perfect explanatory precision (r 2 =0.97).…”

Implementation of a Reconfiguration Algorithm for Cognitive Radio

“…These experiments were run for three iterations using a different random seed. We determined the average effect of a parameter setting on performance, performed an ANOVA on the data, and determined the significant single and multi-factor effects [2], [4], [5].…”

“…The metrics used to evaluate each mutation of the CR's settings included bit loss, latency, and throughput. For more on these techniques see [2], [4], [5].…”

“…In our prior work, we showed via simulation how DOE could be used to identify how different configurations of the CR could improve (or degrade) network performance in simulation [2], [4], [5]. Central to our approach was the use of the DOE analysis to inform the design of the reconfiguration algorithm.…”