Research On Key Techniques Of Spectrum Sensing Based On Cross-Layer Optimization

Cognitive Radio Networks (CRN) has attracted a lot of attention of research in recent years as a promising technique for overcoming the apparent spectrum scarcity problem. In frequency reuse based CRN, the unlicensed users (secondary users) are allowed to utilize the spectrum when licensed users (primary users) are not occupying it. Spectrum sensing to detect the pres-ence of the primary users is, therefore, one of the most important technologies in CRN. In the view of network hierarchy, spectrum sensing can be further subdivided into physical-layer sens-ing and Mac-layer sensing. The physical-layer sensing is main concern on how to make quick and accurate judgments on the channel occupancy status by effective signal detection algorithms, while the Mac-layer sensing is main concern on how to improve the physical-layer sensing effi-ciency by optimizing sensing parameters and strategies.At present, the physical-layer sensing algorithms have a considerable theoretical basis due to the early development of digital signal processing technology. In contrast, the studies on Mac-layer sensing are still in its infancy. Therefore, in supporting and finishing national high-tech exploration-oriented project-’Optimization techniques of wireless resource allocation in cognitive radio networks based on behavior prediction’, we investigate the sensing parameters and strategies in Mac-layer sensing, based on the classical physical-layer sensing algorithms. Specifically, we study the following four issues in Mac-layer sensing, namely sensing period, sensing time, channel selection strategy and the fusion parameters in cooperative sensing. The main contributions and innovations of this dissertation are summarized as follows:(1) In order to improve the detection rate of spectrum resources in multi-channel environ-ment, a multi-channel spectrum sensing period optimization algorithm is proposed. By analyzing the cases which may affect the secondary users to discover or employ spectrum resources in the process of spectrum sensing, we formulate the sensing period optimization problem as a con-strained multi-objective optimization problem, based on the continuous-time Markov chain. In order to avoid the subjective factors on the quality of the solution to this problem caused by tra-ditional "weight method", a relatively complex Genetic Algorithm (GA) is adopted in solving the multi-objective optimization problem. Besides, in order to improve the performance of GA in this application scenario, we also propose an improvement method, which greatly improve both the convergence speed and the quality of the final solution. The proposed algorithm is different from the algorithms in most popular literatures, which set equivalent sensing period to each channel. We give full consideration to the load difference between each licensed channel, and set different sensing periods to them, which greatly improve the detection rate of spectrum resources in multi-channel environment.(2) In cooperative spectrum sensing, previous numerical studies show that the sensing per-formance and the throughput of channel can be improved by increasing the secondary users which are participate in cooperative sensing. Due to the existence of channel capability, the throughput of channel can’t be improved infinitely with the increasing number of secondary us-ers, while the average throughput of secondary users decreases severely. Therefore, maximizing the throughput of secondary users should not be the only objective of CRN, the average throughput of secondary users should also be taken into consideration as another performance metric. In short, it is more reasonable to take the average throughput of secondary users as the evaluation standards in CRN. In the view from above, we jointly optimize the sensing time and the parameters of the fusion rule to maximizing the average throughput of secondary users, based on the k-out-of-N fusion rule. Moreover, the unimodal characteristics of the secondary users’ throughput are proved to be a function of the sensing time for. any given fusion parameter in multi-channel environment. To solve this optimization issue, a cross iterative algorithm is pro-posed. Computer simulations show that when the signal-to-noise ratio of primary user at the secondary receiver is-10dB, compared with the classical fusion rule, the proposed algorithm can gain more than 20% of the average throughput of secondary users.(3) More than an adaptive system, the cognitive radio networks is an intelligent system. The reinforcement learning of the intelligent control theory is adopted to cognitive radio networks, in order to lead secondary users to select the channels, which has the maximum throughput reward with high probability, without any estimation of primary traffic and the dynamic of wireless networks environment. Moreover, a reinforcement-learning-based channel selection algorithm is proposed to increase the throughput of secondary users. The algorithm adopts the interactive online learning technology of reinforcement learning theory, and allocates selection probabilities to the secondary users through times of interaction with the environment and self-learning to in-crease the throughput progressively. Besides, the proposed algorithm also adopts the idea of simulated annealing algorithm to optimize the action of channel selection, so that it can transit from the learning stage to the application stage, smoothly. The simulation results show that compared with the traditional channel selection algorithms, the proposed algorithm can improve the throughput of secondary users significantly, and when the primary traffic statistics are changed, it can attain convergence automatically, which could be an attempt to the future intelli-gent cognitive radio systems.(4) Aiming at the issue of secondary users’energy constraint in cognitive radio networks, energy transmission efficiency is adopted as a metric of the energy efficiency of secondary users. Considering that the unslotted access of primary users may result in collisions with secondary users, spectrum sensing and access strategy of secondary users are modeled and analyzed by continuous-time Markov theory. And an energy-efficient algorithm which jointly optimized the sensing time and access probability is proposed. In this paper, we prove that it does exist an op-timal sensing time which will maximize the energy efficiency of secondary users. In addition, based on the optimal sensing time, an energy-efficient spectrum access strategy is proposed. The proposed spectrum access strategy is different from the traditional access strategies. Secondary users do not directly access the channel which is sensedidle, but calculate the collision probabil-ity with the returning primary user based on the statistical regularity of the idle time of channel and secondary users’ access time, then access channel with probability. The simulation results show that compared with the traditional algorithms which only optimize either the sensing time or the access probability, the proposed algorithm can improve the energy efficiency of secondary users significantly.