Название: Artificial Intelligence in Wireless Sensors and Instruments: Networks and Applications
Автор: Halit Eren
Издательство: CRC Press
Год: 2025
Страниц: 297
Язык: английский
Формат: pdf (true), epub
Размер: 24.3 MB

This book heralds a new era in instrumentation and measurements. It combines Artificial Intelligence (AI) and wireless communications technologies with instrumentation and measurement systems to function as a single unit. AI has advanced considerably due to Deep Learning utilizing artificial neural networks, availability of large and curated datasets, implementation of a new generation of fast processors having millions of transistors in chips, advanced algorithms, competitive commercial interests, and interests of governments to gain advantages. At the same time, new and highly advanced wireless technologies open new frontiers in communication systems, both technologically and in terms of applications aspects. Advanced technologies such as 5G and 6G networks enable easy use of communication systems by billions of people as well as by billions of machine-to-machine systems.

In this book, the communication principles are explained and the implementation of AI on wireless networks is discussed. Many examples are provided. The author discusses instruments and instrumentation networks, modern sensors, and transducers in detail.

AI is the technology humans have created where the machines do not only assist us but also think for us creatively in some cases, excelling humans thinking and reasoning. This book includes a chapter explaining how this is done, backed up with more than 50 figures. The security issues, fairness, efficiency, and social impact and acceptance of AI are highlighted. As explained in this book, AI and wireless communications are changing our lives in many ways, including entertainment, games, social interactions, medicine and healthcare, R&D, automated living, intelligent transport systems, finance and economy, and the Internet of Things.

Genetic algorithms (GAs) are heuristic (informed) search algorithms that reflect the process of natural selection. They are the most widely explored algorithmic models used by evolution-based computation techniques. In evolution-based computing, a set of candidate solutions are generated and then iteratively updated. These techniques involve algorithms such as genetic programming, evolutionary methods, differential evolution, and cultural evolution. GAs rely on the use of selection, crossover, and mutation operators to create successive generations of better adapted individuals. In a sense, GA behaves as an intelligent process of stochastic searches within a specified search space to solve a problem. It can also be used to optimize weights of ANNs.

Particle swarm optimization (PSO) is a nature inspired search optimization method. It has a few different methods applied frequently, such as the ant colony optimization, cuckoo search algorithms, and bat algorithms. PSO algorithm mimic social behavior of fish schooling or flying birds looking for food sources through collaborative work of the population. The exploration of the search space could be slow since PSO calls evaluation function by the number of its population size in each iteration. It can be trapped in local optimum if improper values are assigned to the parameters.

There are multiple methods and algorithms used in different stages of AI. There still are many challenges at each stage. Some AI supporting algorithms and methods serve multiple tasks, some are applicable in various categories, and some belong to a parent method with slight variations. Algorithms are also dependent on the type of computer language selected, such as Python, Java C++, Lisp, R, and Prolog.

DL is an AI technique that mimics the human brain in understanding and coming to conclusions. DL is based on the ANNs theory. The learning models can be supervised, semi-supervised and unsupervised. There is a repertoire of extremely successful DL algorithms, such as the DNN, CNN, RNN, and LSTM. DNNs progressed remarkably over the past few years since 2020. DL covers many fields and subfields of AI, and it has introduced a new meaning to AI. It finds a wide range of applications, such as computer vision, speech recognition, NLP, image classification, communications networks, national security, finance, and many others. DL can combine low level features to form more abstract high-level representations, attribute categories, and novel feature discoveries in the new data. Most DL networks explicitly include feedback loops thus giving them memory capability, and dynamic behavior that allows handling of very large sequences, such as the speech, text, and images.