Sensing Orders in Multi-User Cognitive Radio Networks

Introduction

In an age where the wireless RF spectrum is becoming increasingly congested and scarcer than ever, the concept of cognitive radios has attracted significant attention as a means for enhancing the efficiency of spectrum usage (Hossain, Niyato & Han, 2009). A cognitive radio is a smart radio that can intelligently detect available wireless channels in its vicinity and dynamically configure its transmission or reception parameters so as to make the best use of these spectrum holes. Cognitive radios are capable of communicating on licensed spectrum without causing interference to the incumbent or primary users of the licensed bands, and therefore hold great potential for improving the efficiency of the usage of licensed spectrum that is otherwise poorly utilized due to static frequency allocations.

However, the spectrum that a cognitive radio would be allowed to operate on can be expected to be scattered and heterogeneous in general. In other words, a cognitive radio would need to search over multiple portions of the licensed spectrum, possibly having different bandwidths and primary user characteristics, in order to select the best free channel for its use. Also, these radios are expected to be small devices, often hand-held with limited footprint, and therefore cognitive radios that can simultaneously sense more than one portion of the spectrum quickly, efficiently and reliably would be expensive to build in practice. As a result, each cognitive user (CU) in a cognitive radio network (CRN) would need to have a sensing-order i.e., an order in which it will sequentially sense the different channels until it finds a suitable channel for its communication.

The sensing-order problem in cognitive radio networks relates to finding the sensing orders for the cognitive users in the network that jointly optimize a chosen objective function. For example, if the objective is to minimize the expected sum of sensing times or equivalently, the expected sum of times to transmission, the sensing orders need to be jointly selected in a way that lets the CUs find disjoint and free channels as early as possible in their respective sensing orders. A more popular objective, which is also the objective of interest throughout this chapter, is the expected sum throughput of all CUs, also referred to as the cognitive throughput. When cognitive throughput is used as the objective, each CU needs to find not just a free channel but also a good quality channel i.e., if a channel it senses to be free has a very bad signal-to-noise ratio (SNR), it can skip the channel and continue sensing according to its sensing order with the good hope of finding a better quality channel in the future that will result in a higher effective throughput. Therefore, with cognitive throughput maximization, the sensing-order problem also encompasses determining optimal stopping rules for each CU i.e., a rule based on its instantaneous channel conditions that it can use whenever it finds a free channel to decide whether to stop sensing or continue sensing with the hope of finding a better channel in the future.

One might wonder, “If the different channels are identical except for their primary-free probabilities, isn’t it optimal to just sense the channels in the descending order of their primary-free probabilities?” Jiang, Lai, Fan and Poor (2009) showed that when there is a single CU in the network, this intuitive sensing order is indeed optimal when the CU does not use rate adaptation i.e., when the CU transmits at a constant rate irrespective of the channel quality. However, if the CU uses rate adaptation, the intuitive sensing order is not optimal in general. Finding the optimal sensing order involves computing the cognitive throughput as a function of the sensing order and using a dynamic programming approach to search for the optimal sensing order. The section titled Background briefly discusses these and other results for a single-user CRN that will set the background for the material that follows in this chapter.

This chapter focuses on multi-user CRNs. In multi-user CRNs, expressing the cognitive throughput itself becomes challenging because in addition to the complexities in a single-user CRN, the cognitive throughput also depends on how contentions are managed among CUs, or in other words, how multiple CUs are allowed to access a common set of channels. Based on earlier work (Misra & Kannu, 2012a; Misra & Kannu, 2012b), this chapter considers the following two broad lines of contention management strategies and studies the optimal sensing order separately for each of them.