Plenary session

Abstract

Although cognitive radios has seen an enormous interest from various research communities, arguably even the definition of what it is still not clear. In particular, depending on the person's background the adapted definition varies: For example, RF antenna/reconfigurable hardware community treats it as an upgrade of an software-defined radios (SDR's), PHY/MAC and communication theory researchers consider it to be all about dynamic spectrum sharing (DSS), and computer scientists seem to believe it is primarily a radio with machine learning. However, a more suitable notion of a true cognitive radio would need to encompass all these, plus perhaps some more. With this broad view of developing a radio with true cognitive abilities, we define a cognitive radio as “an intelligent wireless communications device that has the ability to reason and learn from the observed RF environment to self-decide optimal communications mode and can optimally self-configure its hardware to support the selected mode”. This plenary will discuss how the operational architecture of such a cognitive radio might look like. We identify what essential functionalities and components a cognitive radio architecture should possess if it is to live up to a such bold vision of a cognitive radio device. We will argue that many notions of cognitive radios found in current literature truly cannot be counted as cognitive radios: It is difficult to see how they differ from simple adaptive algorithms already found in many wireless communication systems. We present our vision for developing a truly cognitive radio and outline a research agenda that we are pursuing at the University of New Mexico. A true cognitive radio has to be a radio with a brain, called the cognitive engine, that is able to autonomously learn from its past experience and current observations, and self-reconfigure within the boundaries of its real-time reconfigurable, yet constrained, hardware. The talk will emphasize the need for developing sophisticated real-time reconfigurable hardware/antennas and autonomous machine learning algorithms for radio environments, if one is to achieve that goal. We also emphasize that such a design of cognitive radios need not be limited to simple dynamic spectrum sharing that is the focus of almost all existing work on cognitive radios.