Research On Resource Allocation Technology Of D2D Communication Based On Convolutional Neural Network

Device-to-Device(D2D)communication technology has become a hot spot technology for next-generation mobile networks due to its excellent performance in short-range communications.In mobile communication systems,D2 D technology can significantly increase system throughput and achieve high spectral efficiency and energy efficiency.However,the way of sharing cellular network spectrum resources in the D2 D implementation mechanism will bring link interference between cellular users and D2 D users.Thus,resource optimization in D2 D communication scenarios has become the focus in the research of mobile communication networks.At present,the mainstream resource allocation technology with convex optimization as the core can optimize network performance,reduce interference between communication links,and improve network transmission performance,but the related optimization iteration algorithm has high computational complexity,which limits its application in specific communication scenarios.Therefore,it is of great research significance and practical value to study the fast solution algorithm for resource allocation problems in D2 D communication scenarios.The dissertation studies the problem of D2 D communication resource allocation in the multiplexing mode in cellular networks.The main work are as follows:(1)An end-to-end single cell downlink D2 D power allocation algorithm based on convolutional neural network is proposed.In the next generation mobile communication network model,there are three commonly used resource allocation modes in D2 D communication scenarios: dedicated mode,multiplexing mode and cellular mode.The dissertation focuses on the analysis of the complexity of the D2 D resource allocation method in the multiplexing mode scenario.The resource allocation model in this scenario is established,and an end-to-end convolutional neural network-based power allocation method is proposed.The algorithm also considers the imperfect channel state information,user's quality of service and other constraints.Compared with the traditional method of convex optimization power allocation algorithm,it can effectively reduce the computational complexity.(2)Compared with the single-cell system,the communication link and interference of multi-cell system are more complicated,and the fast calculation of the D2 D resource allocation strategy of the multi-cell is also more important.Aiming at the problem of resource allocation in multi-cell communication scenarios,an adaptive Multi-DCNN resource allocation algorithm is proposed.The Multi-DCNN algorithm takes the system spectrum efficiency as the optimization goal,meets the upper and lower power limits and minimizes interference as the constraints to establish the optimization function,and fully considers many constraints such as the best system performance,the smallest interference,and the lowest quality of service.Also,the design of loss functions can satisfy power constraint requirements and the difference in demand,which gives an idea for dynamically adjusting the parameters of the neural network model under the variable scenes.Simulation results demonstrate that the adaptive method proposed in the dissertation is suitable for multi-cell complex scenes and has low computational complexity.In the dissertation,the problem of D2 D communication resource allocation in specific scenarios is studied,the single cell and multi-cell scenarios are considered.The proposed fast solution algorithm for resource allocation problems based on convolutional neural networks has good algorithm performance and low complexity.