| 引用本文: | 沈林之,王少尉.深度神经网络指导下的无线覆盖预测算法.[J].国防科技大学学报,2020,42(4):18-23.[点击复制] |
| SHEN Linzhi,WANG Shaowei.Wireless coverage prediction algorithm under the guidance of deep neural network[J].Journal of National University of Defense Technology,2020,42(4):18-23[点击复制] |
|
| |
|
|
| 本文已被：浏览 8069次 下载 5664次 |
| 深度神经网络指导下的无线覆盖预测算法 |
| 沈林之,王少尉 |
|
(南京大学 电子科学与工程学院, 江苏 南京 210023)
|
| 摘要: |
| 为保证新一代移动无线网络能够根据实时覆盖情况动态地调节小区天线参数,需要实现高效且准确的无线覆盖预测。传统的求解方法通过精确的场强预测判断天线参数的优劣,虽然精度很高但需要大量的计算资源,无法满足5G和后5G移动网络通过实时覆盖预测进行射频参数动态调整的实际需求。现采用基于深度神经网络的算法对给定天线参数的覆盖效果进行预测,以取代对目标区域的精确场强预测。数值结果表明:该方法能够在保持计算准确性的同时显著减少计算量,为5G动态网络规划提供基础性参考数据。 |
| 关键词: 动态网络规划 无线覆盖 场强预测 深度神经网络 |
| DOI：10.11887/j.cn.202004004 |
| 投稿日期：2019-12-25 |
| 基金项目:国家自然科学基金资助项目(61671233,61801208, 61931023) |
|
| Wireless coverage prediction algorithm under the guidance of deep neural network |
| SHEN Linzhi, WANG Shaowei |
|
(School of Electronic Science and Engineering, Nanjing University, Nanjing 210023, China)
|
| Abstract: |
| In order to adjust the parameters of cell antennas dynamically according to the real-time coverage in the new generation mobile wireless network, it is necessary to predict the wireless coverage efficiently and accurately. The traditional solution method is to judge the antenna parameters by accurate field strength prediction in the target area. The method is accurate but wastes large amounts of computing resources, which cannot meet the actual needs of 5G and beyond 5G mobile networks to dynamically adjust the radio frequency parameters through real-time coverage prediction. Here the algorithm based on deep neural network was proposed to predict the coverage under given antenna parameters in order to replace the accurate field strength prediction of the target area. Numerical results show that the algorithm can keep the accuracy of the calculation while significantly reducing the computing resources, which provides basic reference data for 5G dynamic network planning. |
| Keywords: dynamic network planning wireless prediction field strength prediction deep neural network |
|
|
