2.1 Introduction to AI
Artificial Intelligence (AI) is a field of science that is constantly evolving and accelerating. It has recently witnessed great momentum in being one of the scientific fields that have become affecting all sciences. AI has transformed the research path to new directions in order to provide effective solutions to many problems facing all science and engineering fields. In fact, the concepts of AI go back to the 1940s and 1950s, when scientists from different disciplines explored the possibilities of artificial brains and defined machine intelligence.
AI systems can be divided into three types: analytical AI, human-inspired AI, and humanized AI. Analytical AI has characteristics compatible only with cognitive intelligence, where new knowledge and decisions are generated through learning and based on previous experience. Human-inspired AI is a mix of cognitive and emotional intelligence. Human emotions are understood, in addition to the cognitive elements, and then used in determining decisions. Humanized AI uses cognitive, emotional, and social intelligence, capable of being self-conscious, and self-aware in interacting with others.
The basic idea of AI is based on a simulation process of the interaction of data in human thinking, trying to understand human intelligence and then developing intelligent machines. AI has the ability to access objects, categories, their characteristics, and the relationships between them in order to apply knowledge engineering. AI aims to expand the capabilities of mankind in carrying out various tasks and consolidate the principles of intelligence in machines and devices in order to save time and effort and to provide distinguished services in various fields. Nowadays, we are witnessing the emerging of many smart devices in different fields, especially in engineering and medical sciences. Specific examples are computer vision, natural language processing, the science of cognition and reasoning, robotics, game theory, and machine learning (ML). Intelligent machines would have some of the capabilities related to human thinking in dealing with problems and make appropriate decisions for any event that may appear during machine operation.
It is known that existing networks lack the intelligence needed to support future nextgeneration networks that are expected to be self-adaptive. Mobile networks consist of a large number of elements that interact with each other, creating a great complexity in the system that operates these elements together. Wireless networks constitute one of the most important areas that aspire to benefit and consolidate the principles of AI in order to adopt solutions to many problems appeared previously and appear currently in this field. Although we observe a great revolution in scientific research that relies on AI tools to develop and design wireless networks, applying AI approaches to network planning, design, and operations is still in the early stages. This is due to the fact that existing network architectures are not suited to the AI-enabled networks. Researchers are looking not only at the use of AI-based solutions to current problems, but noticeable research have returned to previous problems and tried to develop AI-based solutions. Later in this chapter, we will discuss recent research issues that can benefit from and exploit the principles of AI and ML.
The main research directions that use the AI paradigm are as follows:
Expert SystemsAn expert system is a software system that relies on human expertise for decision-making. It is appropriate to deal with problems that involve incomplete information or big data.
Machine LearningML relies primarily on how the computer simulates the behavior of human learning, then restructures the knowledge and acquires new skills to continuously improve performance.
Pattern RecognitionThe concept of pattern recognition is applied to process monitoring that assumes a relationship between data patterns. The research in pattern recognition includes two main issues: the first relates to object perception and the second relates to determining the category to which the object belongs.
Neural NetworksThe concept of artificial neural networks is based on non-linear mapping between the system’s inputs and outputs. It consists of interconnected neurons arranged in layers. The layers are connected, allowing signals to propagate from the layers’ inputs across the network. A neural network stores data, learns from it, and improves its capabilities to sort new data.
Deep LearningDeep learning is the application of the concept of artificial neural networks to learning tasks that contain more than one hidden layer. It is part of a larger group of ML techniques that are based on representations of learning data. Deep learning concepts come from artificial neural network research, which opened a window to a new field of ML. Concepts of deep learning have been applied in various fields including computer vision, speech recognition systems, natural language processing systems, voice recognition systems, social networking systems, automatic translation systems, and bioinformatics systems, where the adoption of deep learning techniques has led to more effective results as compared to human experience and previous systems.
2.2 Self-Organizing Networks
The primary goal of mobile networks is to connect mobile phone users together as well as to the Internet. Therefore, wireless network operators install large number of base stations or access points in the regions that will be covered. Each base station or access point covers a specific geographical area called a cell. Mobile networks allow users to transparently move between cells via a process called handover. Network users wish that the service provided to them is uninterrupted, whether with regard to the quality of phone calls or the speed with which they surf the Internet.
Self-Organizing Networks (SONs) is an evolving technology used to automate planning, configuration, optimization, and healing of networks. SON is included as part of the mobile networks standards such as such as Long Term Evolution (LTE). The quick evolution in wireless network industry have led to parallel operation of 2G, 3G, 4G, 5G, and emerging 6G networks that need to be managed and controlled with minimal human effort. SON is a promising technology to realize solutions for the control and management of this heterogeneous network regime. The technology suggests a set of concepts to automate network management toward a goal of improving quality of service (QoS) and reduce burdens of networks management on network administrators [1].
2.2.1 Operation Principle of Self-Organizing Networks
With SON, network administrators predefine a set of key performance indicators (KPIs) regarding QoS and other operational functions. Then, the network uses modules and algorithms to self-monitor and optimize its parameters, trying to achieve the predefined KPIs. This is considered as a closed loop control process, by which a network gains understanding of the operation environment and users’ behavior and adapts its parameters accordingly to achieve the intended performance goal, but at same time avoid any misconfiguration of parameters that may lead to service disturbances [2]. In the following subsection, we elaborate more on the features of SONs illustrated in Figure 2.1.
Figure 2.1 SON features.
2.2.2 Self-Configuration
Mobile communications networks are heterogeneous networks comprised of multiple technologies, such as LTE, EDGE, and UMTS. The number of mobile users is incredibly increasing which makes the installation and configuration of base stations a tedious process. Therefore, self-configuration is a process that reduces the time required for these tasks.
Self-configuration provides an initial setup of the network elements. It consists of three stages. The first stage relates to automatic connection to the network, security procedure, and establishing a secure connection between network elements
