A Statistical Method for Reconfiguration of Cognitive Radios

Recent developments in computer technology have enabled radio developers to accomplish in software what traditionally was performed with application-specific integrated circuits. A radio that has the core of its functionality implemented in software is called a software-defined radio. When an SDR has the capability to sense, reason, and dynamically adapt to requirements and environmental change, we call this more capable device a cognitive radio. Many private and public agencies are investing in the promise of CR to improve the utilization of radio frequency spectrum. They envision devices that can sense frequency vacancies and dynamically reconfigure to utilize idle channels. The promise of CR depends on the capability of a radio to change operating frequencies, power, and/or modulation schemes (physical layer flexibility). In addition to this physical layer flexibility, there are a large number of opportunities to capitalize on the interplay of the CR physical layer configuration and other parameters in the radio network protocol stack. At the core of CR functionality is the ability to select from thousands of potential configurations to maximize performance-be it in terms of spectrum use, throughput, or reliability. In this article, we describe a method for selecting from a number of potential configurations to fulfill the communication requirements of a CR network. By using accepted statistical methods, we show how parameters at the physical, data link, network, and application layers interact to affect performance. We build upon this parametric insight with our presentation of a technique for predicting radio performance.