Interferrence Prediction And Resource Allocation Research For Ultra-Dense Networks

The Society has become an information society,wireless network plays an important role in people’s life.In order to cope with the large amount of traffic generated by large number of users,ultra-dense network(UDN),as one of the main ideas of the new generation wireless network,effectively enhances network coverage and regional spectrum efficiency by densely deploying access nodes in each unit area.The ultra dense characteristics of UDN also bring some challenges to wireless communication.The intensive deployment of equipment makes the severe interference between cells a limiting factor for the improvement of network performance.It is urgent to effectively manage and coordinate the interference between cells.However,in the traditional cellular system architecture,radio resources are allocated by each cell,which lacks an effective way to solve this problem.The serious interference problem in UDN seriously restricts the capacity gain of the system,and the main bottleneck is the inability to obtain accurate interference information.Big data in wireless network actually contains rich radio interference information.In this paper,a neural network prediction algorithm of uplink interference based on interference model is proposed.This algorithm designs neural network based on interference model,uses big data for training,mines inter link signal-to-interference ratio(SIR)information in wireless network,and realizes accurate prediction of signal-to-interference-plus-noise ratio(SINR)without increasing radio resource overhead.The simulation results show that when the target user’s training data reaches 5000 entries,the average root mean square error(RMSE)of SINR prediction is less than 0.5dB,which effectively reduces the requirements of data volume and computing power,and is suitable for practical applications.Considering the problem of user mobility in the actual environment,the interference relationship will also change,so an effective interference recognition algorithm is proposed to solve this problem.the mobile adaptive interference recognition technology based on progressive training.The parameters in the interference model are gradually trained with the latest runtime wireless data.According to the simulation results,the accuracy of the proposed algorithm is higher and the time complexity is lower than the contrast algorithms,it can keep the average RMSE predicted by SINR lower than 0.5dB in the user moblility scenario,and there is no need to retrain the model,and it has good real-time performance.Further,with the mobile adaptive interference identification algorithm as the core,with the help of reinforcement learning,joint radio resource management(RRM)is carried out among multiple cells to meet the challenges brought by the complexity of 5g system core technology and the flexibility of network deployment to ridio resource allocation and interference management.The mobile adaptive module is used to model the latest interference state of the environment,give high-precision feedback to the reinforcement learning network,and input the interference information into the reinforcement learning network,so as to achieve satisfactory network performance with less training time.