Toward an AI-Enabled SDN-based 5G & IoT Network

Artificial Intelligent technology has been applied to a wide variety of applications including products, systems, and services of the Information & Communications Technology (ICT) and non-ICT or traditional industries. The benefits of these applications includes performance improvement, optimization, intelligent.The 5G mobile/wireless networks have intelligent features of network slicing and edge computing because of network equipment system vendors apply AI technology to the mobile systems. On the other hand, many traditional industries have benefits from AI technology in particular Machine Learning (ML) and Deep Learning (DL). Recently in the agriculture, healthcare, finance, and many other applications and services have adopted AI/ML/DL technology even with the integration of 5G and Internet of Things (IoT). This article focuses on the system architecture and design of open networking (ON) solution of 5G, the approach of SDN/NFV-based 5G and IoT and how AI/ML interact with 5G/IoT and learns from these. We call this interaction as SDN-based 5G/IoT Network AI or AI-enabled SDN-based 5G/IoT Network.


Introduction
Artificial intelligent (AI) is a great technology and applications for which existing solutions require better automation or optimization and there is no good solution by using traditional approaches. So, when people say AI in Network really means using AI techniques to assist network operations, configuration, and management or to play a supporting role. On the other hand, Network with machine learning (ML), deep learning (DL) and other AI techniques used to manage network to be predictive and proactive, then AI/ML/DL becomes the driving force of network operations and management or the enabler of network operations. We call this kind of network is an AI-enabled Network (Yao et al., 2018;Zheng et al., n.d.). AI is also an interdisciplinary science with multiple approaches but the advancements in ML and DL are creating a paradigm shift virtually every sector of the technology industry.

Open Network and 5G
It has been a paradigm shift in the network industry, from a proprietary networking with vendor-dependent soft-ware/hardware to vendor-independent open source and white box switches as illustrated in Figure 1 Figure 2 shows the definition of Internet of Things (IoT) and some examples Until now, most of the systems of Internet of Things (IoT) applications snd services are relying on 4G/LTE or even 3G as the gateway to transport data to the destination for computing and applications creation and execution. Figure 3 illustrates wireless sensor network (WSN) which collects data and/or signals from devices and use 4G/LTE and 5G as the gateway to network and services platform for further application creation. Its equivalent 4 Layer Model of IoT is shown in Figure 4 (Akyildiz, 2017). Those 4 layers are layer 1: sensing and identification, layer 2: network construction, layer 3: information processing and layer 4: integrated application.

SDN-Based 5G
In this article SDN-Based 5G means 5G network is comstructed and deployed by using open source code related to SDN/NFV to do the softwarization & virtualization of 5G. The network architecture of 5G consists of ORAN and virtual EPC (core network). ORAN is an open source radio acess network subsystem which separates RRH/RRU (remote radio head/remote radio unit) from BBU (baseband unit) and its core network is an available open source EPC. Figure 5 illustrates the open network architecture.of 5G. This open 5G network architectre creates the opportunity to be multi-vender to avoid lock-into one vender with proprietary solution.Also the separation of ORAN and EPC may create the new eco system of 5G industry.

Relationships among Al, ML Neural Networks and DL
This section discusses the definitions of AI, ML, DL and their relationships and differences. AI is algorithms, programs, coding can make machine, objects and things more intelligence through learning and/or training. ML is a subset of AI while DL is a subset of AI and ML.
In practical AI system, these technologies are often supplemented by traditional software coding techniques to manage the AI system, to preprocess data used to drive the AI system, and to implement output a to adapters to effectively use the decision or recommendation output. Many forms of AI incorporate ML to enable the AI systems to better adapter to a complex potentially dynamic environment ML involves training or data acquisition that can modify machine behavior and comes many forms: supervised learning, unsupervised learning, reinforcement learning, and online learning.
Neural networking is a specific class of AI ML systems neural networks have shown promise in addressing complex data relationships. DL generally applies to large neural network with thousands of hidden layers wherein training occurs on each layer within the hidden nodes of a neural network. In the recent years deep learning neural networks have become the most promising approach to AI.

AI in Network vs AI-enabled Network
There are two ways to migrate AI software to the existed system. One way is try to use AI algorithms or techniques to assist the system to become more automation, more optimization, and better service. So, AI in network means what AI can assist network to complete network functions. The other way is using AI/ML/DL techniques to complete the network migration. So, AI-enabled Network means AI driven network operation and management. The characteristics of 3 approach to architect and design the network include: Traditional Networking without AI, AI in Network, AI-enabled Network are in the following: 1) Traditional Networking without AI/ML/DL: reactive, no intelligence, no optimization, error prone, inefficient 2) AI in Network is: semi-active, partial intelligent, local optimization, semi-automated 3) AI-enabled Network is: proactive, intelligent, global optimization, automated, scalable Table1 shows the comparison of 3 approaches.

AI-enabled SDN/NFV 5G and IoT Network
The status of open source for 5G network architecture (Lin, Figure 7 illustrates the simplified 5G network system architecture-based IoT applications. SDN-based 5G/IoT AI-enabled is shown in Figure 8. 5G/IoT AI-enabled can be used to manage 5G networks (Nelson, 2020) and IoT applications.

Global AI Industry
The AI market for AI/ML/DL includes software embedded inside chip, device, module, network infrastructure, subsystem, information processing and integrated application. AI may be used in or applied to one of many industries and play as a tool of assistance. AI may also can be a kind of product. And we expect more products are coming out in the global market based on the amount of investment of AI R&D activities in US, China, Europe, Japan, and Taiwan. At last. AI can be as a service (AaaS). We can see more and more AI plays as services and products by itself. AI will become one of the most promising high technology industries.

Toward an AI-enabled SDN-based 5G &IoT Network
This section we focus on the architecture, design, and challenge issues. Figure 6 and 7 have shown the simplified architecture of SDN-based 5G/IoT: Figure 6 without AI and Figure 7 with AI/.ML. The intelligent agents can be applied to each layer of four-layer model in Figure 4. In this case, it applies to network infrastructure & construction layer. Figure 7 illustrates AI/ML has been applied to information processing and integration application. Together with Figures 4, 5, 6, 7 and 8, the architecture of AI-enabled SDN-based 5G/IoT network is shown in Figure 9.

Design and Application of AI-enabled 5G/IoT Network
The design of SDN-based AI-enabled 5G/IoT based on the architecture of AI-enabled 5G/IoT of Figure 9. To ensure the global level optimization, both 5G and IoT have to go through AI/ML/DL learning process.
For the applications of AI-enabled 5G/IoT Network, the fusion of AI, 5G, and IoT creates the new opportunities of smart agriculture or farms, smart healthcare, intelligent transportation, smart city. Figure 10 shows the process and applications created by the integration of 5G, AI, and IoT. 2) The software quality of open source code 3) The completeness of open source for 5G (Snow & Parulkar, 2016) 4) More successful stories for field trials and network deployment needed 5) AI-enabled to identify assistant role or proactive role

Future Work & Conclusion
This article addresses a variety of industries having been AIized by applying AI/ML/DL techniques to act as a tool or to show as an assistant to enhance the other industries. It is not clear what AI industry means. When they are more AI/ML/DL technology used to create or produces more differentiable and unique AI-based products and applications/services, then AI will have its own industry, AI Industry will be one of competitive technology industries.