Research On Spectrum Sensing And Allocation Algorithm For Cognitive Wireless Sensor Networks

With the rapid development of wireless sensor network(WSN)applications,the unlicensed ISM frequency band intensively used for work is becoming more and more crowded,which can no long meet the requirements of WSN low-latency and high-data applications' transmission.The research on Cognitive Radio Sensor Network(CRSN)combining Cognitive Radio(CR)technology with WSN has become a research hotspot.Spectrum sensing and spectrum allocation technology are the key technologies for CRSN to efficiently utilize spectrum resources.In terms of spectrum sensing,spectrum sensing technology based on energy detection does not require prior information of primary users' signal and has low computational complexity.It is very suitable for CRSN,but its detection accuracy of exploring spectral holes can be reduced under the influence of noise fluctuations and shadow fading;in terms of spectrum allocation,reinforcement learning provides an efficient solution for spectrum allocation,however,high-dimensional action state space and Secondary users access,it is necessary to use a more efficient reinforcement learning algorithm with faster convergence speed for spectrum allocation.Aiming at the above problems,this article designs an efficient cooperative spectrum sensing and spectrum allocation scheme for distributed CRSN to achieve the goal of more fully utilizing spectrum resources and reducing communication interference.The main work of the article is as follows:1.Aiming at the problem that the traditional energy detection method is affected by noise fluctuation and shadow fading and reduces the detection accuracy,this article proposes a dual-threshold multi-point cooperative energy detection algorithm that combines wavelet denoising decision-making.Process,the dual thresholds can be adaptively adjusted with the noise uncertainty,and then use the wavelet denoising decision to determine the signal in the fuzzy area between the dual thresholds,and improve and propose an adaptive threshold and threshold function,which improves the denoising decision.Finally,the cooperative detection of multi-cognitive nodes is used in CRSN to improve the accuracy of the detection system under the condition of shadow fading.Compared with the traditional energy detection method,the improved algorithm can ensure the detection accuracy under the influence of noise and shadow fading.1.Aiming at the problem of unstable detection accuracy caused by the influence of noise fluctuation and shadow fading in traditional energy detection methods,this article proposes a dual-threshold multi-point cooperative energy detection algorithm fused with wavelet denoising decision-making.Firstly,a dual-threshold detection process is introduced on the basis of the single-threshold energy detection method,and the dualthreshold can be adaptively adjusted with the noise uncertainty;then the wavelet denoising decision is used to determine the signal in the fuzzy area between the dualthresholds,and an improved method is proposed.The adaptively changed threshold and threshold function effectively improve the accuracy of denoising decision-making.Finally,the multi-cognitive node cooperative detection is used in CRSN to improve the detection accuracy of the system under the condition of shadow fading.After simulation verification and comparison,the improved algorithm can ensure the accuracy of detection under the influence of noise and shadow fading2.Aiming at the joint problem of single-node spectrum access and power control in CRSN,this paper proposes a single-user spectrum access and power control algorithm based on Deep Q-learning Network(DQN).For the purpose of resource utilization and reducing node energy consumption,the joint problem of single-user spectrum access and power control is mathematically modeled as a Markov Decision Process(MDP)model,and then on the basis of this MDP model,Based on the DQN algorithm,the solution is effectively solved,and the simulation results show that the method is effective in terms of convergence,spectrum resource utilization and energy consumption,which improves the spectrum resource utilization rate of the system and prolongs the node lifetime.3.Aiming at the problem of optimal allocation of spectrum resources jointly by multiple sub-users in CRSN,considering the influence of noise,sub-user location and path loss on resource allocation and communication performance,a long short-term memory deep Q network(DQN+LSTM)distributed spectrum allocation strategy,in order to maximize SNIR and maximize the access success rate,the reward function and distributed spectrum access scheme are designed,each user has its own DQN+LSTM network training and learning The best channel selection strategy is verified by simulation and compared with the DQN and Q-learning algorithms.the effectiveness of the algorithm.