基於機器學習進行網路切片的辨識

Abstract

中文摘要 隨著無線通訊的蓬勃發展，目前擁有大頻寬、高傳輸、高擴充性的5G時代及未來的6G時代，使得人們在使用無線通訊網路的依賴性與日俱增，相較於4G-LTE不僅提升網路速度在頻寬上也更大，當然僅只有這些演進無法彰顯5G-NR的強大。5G-NR雖然在網路速度、頻寬、設備的連結性數量相較4G-LTE有大幅的提升但是網路資源依然是有限的，所以在5G網路裡有了網路切片的產生，何謂網路切片呢?網路切片主要的功能是將無線網路的資源透過虛擬化的方式將其功能進行分割，而目前根據3GPP在TS 28.541中提出了5項切片服務為URLLC(Ultra-Reliable and Low-Latency Communications) 、eMBB(Enhanced Mobile Broadband)、MIoT(Massive Internet of Things)、V2X (Vehicle-to-Everything)、HMTC(High Performance Machine-Type Communications)。 本次專題所使用的切片類別有URLLC、eMBB、MIoT將所需要的無線網路服務進行這三種分類，當輸入層收到了16項參數會進行網路切片類型的選擇，再透過tensorflow建立一個ANN(Artificial Neural Network)的模型，模型訓練完後模型的輸出層會選擇出應用適合哪項網路切片的類型，最後並進行資料驗證及比對來做正確率的分析。

Abstract With the booming development of wireless communications, the current 5G era with large bandwidth, high transmission, and high scalability and the future 6G era have made people more and more dependent on the use of wireless communication networks, which not only improve the network speed and bandwidth compared to 4G-LTE, but of course, only these evolutions cannot show the power of 5G-NR. Although 5G-NR has significantly improved the network speed, bandwidth, and number of connected devices compared to 4G-LTE, the network resources are still limited, so network slicing is created in 5G networks. What is network slicing? The main function of network slicing is to partition the resources of wireless network by virtualization, and according to 3GPP, five slicing services are proposed in TS 28.541: URLLC (Ultra-Reliable and Low-Latency Communications) According to 3GPP in TS 28.541, five slicing services are proposed: URLLC (Ultra-Reliable and Low-Latency Communications), eMBB (Enhanced Mobile Broadband), MIoT (Massive Internet of Things), V2X (Vehicle-to-Everything), and HMTC (High Performance Machine-Type Communications). The slicing categories used in this project are URLLC, eMBB, and MIoT. These three categories are used to classify the required wireless network services. After the model is trained, the output layer of the model will select which type of network slice is suitable for the application, and finally, the data will be verified and compared to do the correctness analysis.