Research On Interference Prediction And Resource Allocation Of Mobile Communication Network Based On Intelligent Algorithm

In response to the explosion in demand for wireless networks,5G uses ultra-dense networks(UDN)solutions.Intensive deployment of network equipment can greatly increase the capacity of wireless networks.However,increasing the deployment density will inevitably bring serious intercell interference.The severe inter-cell interference in 5G puts forward higher requirements for inter-cell radio resource management technology.Unfortunately,existing solutions cannot handle such complex scenarios due to their mechanical limitations or requests for unavailable information.Therefore,it is of great significance to study the enhanced evolution technology of 5G system.Among them,wireless interference sensing and radio resource management enhancement technologies are particularly important.Therefore,the main contributions and innovations of this paper include:A resource allocation framework based on machine learning is proposed,which is composed of a performance inference module and a resource allocator module.Combined with this framework,an interference recognition algorithm is proposed which can accurately identify interference relationships in wireless networks.A high performance QoS-driven resource allocation algorithm is also proposed.The resource allocation algorithm based on enhanced learning is proposed to avoid ICI in UDN.At the same time,it maximizes the total effective capacity of the network while satisfying the basic QoS requirements.Each innovation point of this paper will be introduced in detail below:(1)An intelligent and accurate interference identification technology based on 5G wireless big data and XGBoost machine learning algorithm is proposed.The up-SINR prediction model was trained by XGBoost regression prediction algorithm.Then,when any interference relation is input into the prediction model,the user’s UP-SINR under the interference can be predicted.Thus,accurate and fine-grained network interference matrix can be obtained efficiently.The prediction results are applied to the subsequent design of wireless resource allocation algorithm.This solution is simple to implement and more similar to the actual network scenario.Meanwhile,real-time,high efficiency,high precision and complete interference prediction are realized.Meanwhile,the performance of linear regression algorithm is poor,and the learning time of neural network algorithm is too long.XGBoost algorithm improves the prediction error performance by 80%and 45%respectively.It realizes the amount of second level training data and sub-second level training time when the prediction performance meets the requirement of 0.5dB.(2)An intelligent uplink wireless resource allocation algorithm based on user quality of service(QoS)is proposed.This algorithm introduces the concept of effective capacity EC,which can model the wireless channel according to QoS metric,including data rate rk,delay dmax and delay violation probability ε.Therefore,it can well reflect QoS requirements.Meanwhile,the algorithm is based on deep reinforcement learning(DRL)and has carefully designed formulas corresponding to states,actions and rewards.Specifically,for each RB allocated,the algorithm runs on the same number of agents as the base stations.On each run,the algorithm determines whether RB should be assigned to the base station and,if so,to the user associated with the base station,along with a specific power value.Thus,a compact deep Qnetwork(DQN)reduces computational effort during training and evaluation.At the same time,the availability of the algorithm is improved.Simulation results show that the proposed algorithm can effectively use the obtained interference diversity to allocate the same RBs to users with less mutual interference,thus significantly reducing the co-channel interference and thus improving the SINR and effective capacity value of the system.