Sequential Spectrum Sensing In Cognitive Radio Networks

Nowadays the wireless communication networks display the traits of broadband and service diversity. There are more new services demand for higher capacity and lower latency. However in conventional wireless networks, radio spectrum resources are statically allocated to the licensed users, which caused both spectrum scarcity and spectrum underutilization. To alleviate this problem, FCC suggested that the Cognitive Radio (CR) is a promising technology to improve spectrum utilization efficiency. In cognitive radio networks, the cognitive users can find the free channels which are not occupied by licensed users through spectrum sensing, and then decide in which one to access. Spectrum sensing and spectrum decision are two fundamental functions in spectrum access, so optimization of spectrum sensing and spectrum decision is one of the hot research topics in cognitive radio networks.A joint spectrum sensing and decision optimization scheme named sequential spectrum sensing is discussed in this thesis. In sequential spectrum sensing, there are two problems should be solved:1. in what order should the channels be sensed, i.e. channel sensing ordering and2. when to stop sensing and start transmission, i.e. the stopping rule. Firstly, this paper deals with one user case and designs sequential spectrum sensing methods for various services. And then for multi-user networks, both distributed and centralized sequential spectrum sensing methods are proposed. Finally, sequential spectrum sensing schemes are designed for the case when there’s no any prior knowledge of channel statistics. The main contributions of this paper are as follows:Firstly, the advantage of sequential spectrum sensing scheme compared to traditional spectrum sensing and decision schemes is demonstated. Two traditional spectrum sensing and decision schemes are considered:a step-by-step optimization scheme and a Partially Observable Markov Decision Process (POMDP) based scheme. The efficacy of these schemes is studied by both theoretical analysis and simulations. The results show that the step-by-step optimization scheme can find the largest spectrum opportunity, but has the lowest slot effectiveness. And the POMDP scheme can get the highest slot effectiveness, but the lowest spectrum opportunity. However, the sequential spectrum sensing scheme can make a good balance between spectrum opportunity and slot effectiveness. It has the best performance in transmission capacity and energy efficiency.Secondly, different sequential spectrum sensing methods are proposed for various applications. The conventional researches failed to fully consider the QoS demands for different services. In this paper, a new greedy search algorithm is proposed to minimize the spectrum access delay for real-time applications. The probability that a channel meets the sustainable rate constraint is considered in designing the channel sensing order. The simulation results show that this new sequential spectrum sensing method can get smaller spectrum access delay than traditional ones. For best-effort applications, the aim is to maximize the transmission capacity. Both greedy search and dynamic programming are proposed to find the optimal channel sensing order. Results show that the proposed method can get more throughput than traditional ones. After getting the optimal channel sensing order, the optimal stopping rule is obtained by a backward recursive method, and a suboptimal stopping rule:1-Stage Look Ahead (1-SLA), is proposed to reduce the computation complexity. Since implementing stopping rule needs a probing phase to scan the SNR of the channels, which leads to more time and energy cost. So the efficacy of stopping rule is verified, and the results show that with larger channel numbers and smaller channel probing time, it’s better to adopt the stopping rule.Thirdly, new sequential spectrum sensing methods are proposed for multi-user cognitive radio networks. The channel availability, channel achievable rate, multi-user diversity and collisions among CR users should be considered comprehensively in multi-user case. This paper investigates the sequential spectrum sensing problem for both distributed and centralized network scenarios. For the distributed scenario, a novel potential function is proposed to represent the two-way selection index between users and channels, which can fully exploit the multi-user diversity. For the centralized scenario, a sensing matrix is proposed to represent the channel sensing orders. The throughput loss due to collision is calculated, and a modified greedy search algorithm is proposed to obtain the optimal sensing matrix. Compared with traditional methods, both the new distributed and centralized schemes can get larger throughput, lower collision and better fairness. Finally, it’s found that in multi-user case, the sensing-order setting and the stopping rule should be jointly designed from a systematic point of view. So an online sequential spectrum sensing scheme is proposed. It can adjust the sensing order online according to the real-time sensing results and stopping action. But an information exchange phase is needed to implement this scheme. The results show that when the time cost of the information exchange phase is relatively small, this online scheme performs better than both distributed and centralized schemes.Finally, some new sequential spectrum sensing methods are proposed when the information about channel statistics are not available for cognitive users a priori. The need to learn this information online creates a fundamental trade-off between exploitation and exploration. This can be formulated as a Multi-Armed Bandit (MAB) problem. Here we only consider one user case, and the sensing-order setting and stopping rule are discussed separately. For sensing-order setting problem, the regret of classic UCB1policy will exponentially increase with the number of channels. Thus this paper proposes two new methods:UCB1with virtual sampling and UCB1index based greedy search algorithm. The UCB1with virtual sampling method fully explores the dependencies between different sensing orders, which increases the learning efficiency of the user. The UCB1index based greedy search method modifies the potential function of classic greedy search algorithm with the UCB1index, which makes the decision converge to the optimal channel sensing order rapidly. The simulation results show that both the UCB1with virtual sampling and UCB1index based greedy search methods outperform traditional ones. For the stopping rule setting problem, based on classic UCB1policy, two new methods are proposed: UCB1with virtual sampling and modified UCB1index method. The principles of both UCB1with virtual sampling methods are the same:explore the dependencies between different arms. The modified UCB1index method can reduce the space of the candidate arms and increase the convergence rate. It is shown that compared with conventional methods, both new methods can get much lower regret and much higher convergence rate.