Nonita Sharma - Cyber-Physical Systems: A Comprehensive Guide

R. Sharma and N. Sharma

Dr. B.R. Ambedkar National Institute of Technology Jalandhar, Punjab, India

DOI: 10.1201/9781003202752-1

Helen Gill of the National Science Foundation coined the term Cyber Physical System (CPS) in 2006 []. In addition, CPS aims to improve the system's functionality, reliability, usability, and safety.

The industry is now witnessing the fourth Industrial Revolution in CPS []. Other than advancement in technology in CPS, these devices work without human intervention and can communicate with each other and produce huge amounts of data. This data is analysed and different services are provided to the user.

CPS is also called an industrial application of IoT that can control physical objects with the help of cyberspace []. The concept of industrial IoT has emerged with CPS that provides machine-to-machine communication without human intervention. With this automation, a machine can work with its full potential and also offer optimized services. Different sensors and actuators are used in these systems to make them autonomous.

As CPS systems become more commonly utilised, the number and quality of attacks is increasing. Protecting these systems from different attacks is challenging in a human-free environment []. If proper security measures are not taken into consideration, it is easy to attack the system. So, it is a concern of prime importance to secure the design and provide a good service simultaneously.

This chapter demonstrates the application of artificial intelligence (AI) in CPS. The following section introduces the concept of CPS to readers, followed by the introduction of AI, and the application of AI in CPS.

CPS is an interconnection of a physical entity with communication and computation infrastructure [6,]. CPS focuses on networking several embedded devices so that they can work together. There are different embedded systems such as smartphones, cars, and other devices; without these devices, we cannot imagine our modern life. All of these devices cannot be accessed remotely. With the help of CPS, these devices can be integrated and communicate with each other and can be accessed remotely. Connected car technology is an example of CPS; these days, most cars come with this technology and it helps us remotely turn on and off the car engine, check the location of the vehicle, and turn on the air conditioning before we reach the car. It'd be perfect if one could switch on the home air conditioning on the way back from work, bringing the indoor room temperature back to normal. Or an electric kettle could start making tea while the person is still in bed in the morning. System monitoring may also benefit from remote access to process data. The details obtained from remote diagnostics assists service members in bringing the appropriate instrument and replacement component. The device will order replacement parts on its own, using the corresponding network infrastructure. Also, now, there are many areas of use for CPS, such as surgical devices, driving machines, etc.

]. These systems need a networking interface to share data with other embedded systems or the cloud. The most critical aspect of a CPS is data exchange, which allows data to be exchanged. The Internet-enabled CPS is often referred to as the IoT. In other words, a CPS is a network-capable integrated device that can send and receive data [6].

According to John McCarthy [], the father of AI, artificial intelligence is the science and engineering of creating intelligent machines, brilliant computer programs. AI is a field of computer science that deals with creating programs that exhibit intelligence. There are two forms of AI: narrow and universal AI. Many of the features of human intelligence would be present in ubiquitous AI, such as the ability to comprehend words, distinguish between objects and sounds, erudition, and problem solving. On the other hand, narrow AI shows certain aspects of human intelligence and can perform certain functions exceptionally well but lacks in other fields. A software that can recognize photographs but not anything else is an example of narrow AI.

To achieve the idea of AI, different algorithms and models are used. These learning algorithms and models are called machine learning (ML). ML is a learning process that is used to make a machine intelligent. Different learning methods are used to make machines intelligent, and they are broadly classified into three categories: unsupervised, supervised, and reinforcement learning. Supervised learning is learning where a device is provided with the label data means the machine knows for a specific input and its output during the training phase. In the testing phase, the machine is given data where the label is provided only on input data and it tries to predict the output label.

On the other hand, in unsupervised learning, there is no label on output data during any phase, machine try to find out hidden features in the data. Reinforcement learning works on the concept of agent and environment where agents learn from the surrounding environment and try to improve their performance. On the basis of learning, there are different techniques that are not limited to: decision tree, K-nearest neighbour (KNN), random forest, support vector machine (SVM), logistics regression, and artificial neural network (ANN). ANN is the technique that is inspired by the human neural system. In the human neural system, there are millions of neurons that work together to tmke a decision in same manner when we use a number of artificial neural networks to make an intelligent machine is called deep neural network and a new term to come into the picture is known as deep learning (DL). DL is an extension of ML that has the same task to make a machine intelligent, but in DL, the number of artificial neurons are very large. Recurrent neural network and convolution neural network are some of the most popular DL models. Applications of AI are used in different domains to improve their performances and user experiences: web search engine, spam detector, image recognition, voice recognition, and sensory data analysis are some of the applications in which tremendous improvement has been done with AI.

A CPS uses computer-aided algorithms to track and manipulate physical elements that are capable of manipulating and reacting to their physical environment, using a mixture of machine sensors, integrated computer intelligence, and multiple communication mechanisms. With rapid advancements in AI and connectivity, demand for CPS is increasing, such as connected and autonomous vehicles that track and interact with their environments, and smart devices that maximise energy usage depending on the atmosphere and the occupant's conduct.

A CPS is becoming data-rich, allowing new and higher degrees of automation and autonomy. For making this automation possible, AI plays a very important role. AI provides the result with great accuracy and precision. AI not only is used to make the system automated and fast but it is also used to secure the system and provide privacy.

A CPS is focused on information-processing computer systems installed in goods such as vehicles, aeroplanes, and other equipment. A computer machine is used to execute a complex operation. The brakes of a car, for example, are controlled by an integrated ABS system (anti-lock braking system). Since a CPS contains both cyber and physical elements, it is referred to as a cyber-physical device. To put it another way, a CPS is a system that combines digital and physical controls. The components of cyber-physical systems, such as networked control systems (NCS) and feedback systems, are distributed and interconnected through communication networks. Cyber-based manufacturing, smart grids, water, and wastewater networks are examples of NCS. Data moving between various areas of an NCS may be vulnerable to a variety of attacks.