Research On The Algorithms Of Dynamic Spectrum Access And Spectrum Resource Allocation In Cognitive Wireless Sensor Networks

As the sensing,processing and communication devices become smaller and less expensive,cognitive wireless sensor networks(CWSNs)are more practical and will be constantly updated.However,the problems of spectrum resources shortage and energy causality limit the development of cognitive wireless sensor networks.Aiming at solving these problems,our paper starts the research from the dynamic spectrum access algorithm and the spectrum allocation algorithm.The main work and innovation of this paper are summarized as follows.1.Aiming at solving the problems of low utilization rate of important experiences,large energy consumption and slow convergence in traditional dynamic spectrum access based on deep Q-learning,a dynamic spectrum access algorithm for CWSNs based on prioritized experiences replay deep Q-learning is presented in this paper.Firstly,to make full use of reliable experiences so as to improve the convergence speed,we firstly change the way that the deep Q network breaks the correlation among samples by random experience replay as by prioritized experience replay.Secondly,to achieve lower consumption and lower memory in the update of memory bank while ensuring the performance of the system,a batch deletion method is presented for unnecessary experiences in memory bank.The simulation results show that the algorithm in this paper is superior to the spectrum access algorithm based on double deep Q-learning in terms of extending the life cycle,reducing the congestion probability and improving the throughput of secondary user nodes.And compared with the traditional random spectrum access algorithm,this algorithm reduces the congestion probability by 6%-10%and increases the throughput by 18%-20%.2.Aiming at improving the energy and spectrum efficiency under the energy causality and the collision causality,a spectrum resource allocation algorithm for CWSNs based on network coding and an energy-harvesting-first model is proposed in this paper.It firstly applies network coding technology to transmit data in order to balance the energy consumed by nodes and to reduce the probability of interruption leaded by overloading.Then,to allocate spectrum resource among spectrum sensing,radio-frequency energy harvesting and data transmission,an energy-harvesting-first model is established according to the energy state and transmission load of nodes,and the optimal sensing time is calculated with the goal of maximizing throughput.The simulation results show that the algorithm in this paper reduces the energy outage probability by 5.3%-10.1%and 2.25%-4.5%compared with the static resource allocation and joint optimization resource allocation of a and ? respectively.Furthermore,it also outperforms the two state-of-the-art algorithms in terms of average throughput.