Edge of Intelligence (eBook)

Exploring the Frontiers of AI at the Edge

Shubham Mahajan, Sathyan Munirathinam, Pethuru Raj (Herausgeber)

eBook Download: EPUB

2025
658 Seiten
Wiley-Scrivener (Verlag)
978-1-394-31438-6 (ISBN)

Infos

The book offers cutting-edge insights and practical applications for Edge AI, making it essential for anyone looking to stay ahead in the rapidly evolving landscape of artificial intelligence and Edge computing.

Edge of Intelligence: Exploring the Frontiers of AI at the Edge examines the transformative potential of edge AI, showcasing how artificial intelligence is being seamlessly integrated with Edge computing to revolutionize various industries. This book offers a comprehensive overview of the latest research, trends, and practical applications of Edge AI, providing readers with valuable insights into how this cutting-edge technology is enhancing efficiency, reducing latency, and enabling real-time decision-making. From optimizing vehicular networks in the era of 6G to the innovative use of AI in crop monitoring and educational technology, this book covers a broad spectrum of topics, making it an essential read for anyone interested in the future of AI and Edge computing.

Featuring contributions from leading experts and researchers, Edge of Intelligence highlights real-world examples and case studies that demonstrate the practical implementation of edge AI in diverse sectors such as smart cities, recruitment, and nano-process optimization. The book also addresses critical issues related to privacy, security, and the fusion of blockchain with edge computing, providing a holistic view of the challenges and opportunities in this rapidly evolving field.

Audience

Engineers, data scientists, IT professionals, researchers, and academics in the fields of artificial intelligence, computer science, and telecommunications, as well as industry professionals in sectors such as the automotive, agriculture, education, and urban planning industries.

Shubham Mahajan, PhD, is an assistant professor at Amity University, Haryana with a remarkable track record in the field of artificial intelligence and image processing. He has published over 77 articles in peer-reviewed journals and conferences, as well as eleven Indian, one Australian, and one German patent. His research includes video compression, image segmentation, fuzzy entropy, nature-inspired computing methods, optimization, data mining, machine learning, robotics, and optical communication.

Sathyan Munirathinam, PhD, is a senior manager on the Customer Service Data and Diagnostics team for the ASML Corporation with over 24 years of experience in business intelligence and 17 years in the semiconductor industry. His responsibilities involve developing and executing a roadmap for data and diagnostics innovation for customer service engineers, aiming to transition equipment from unscheduled downtime to scheduled maintenance. In addition to this role, he has authored numerous papers and participated in numerous international conferences.

Pethuru Raj, PhD, is a chief architect in the Edge AI division of Reliance Jio Platforms Ltd., Bangalore. with over 23 years of IT industry and 9 years of research experience. He has been granted two international research fellowships from the Japan Society for the Promotion of Science and the Japan Science and Technology Agency. His research interests include the industrial Internet of Things (IIoT), efficient, explainable, and Edge AI, blockchain, digital twins, cloud-native and edge computing, green and generative AI, and quantum computing.

Textauszug

1
A Review on Computational Optimization Strategies and Collaborative Techniques of Vehicular Task Offloading in the Era of Internet of Vehicles and 6G

Aishwarya R.*, V. Vetriselvi and Meignanamoorthi D.

Department of Computer Science and Engineering, Anna University, Guindy, Chennai, Tamil Nadu, India

Abstract

The Internet of Vehicles (IoV) and emerging 6G communication technology have recently advanced, empowering intelligent vehicles to support pervasive services while also providing an efficient and convenient driving experience. Furthermore, massive amounts of data are being generated by vehicular applications. The in-vehicle computing capability is insufficient to meet vehicular applications’ time-sensitive and computation-intensive demands. In such a scenario, task offloading towards other resource-rich computing devices can be considered to process vehicular tasks, thereby improving the application’s Quality of Services (QoS). In this paper, a comprehensive review of task-offloading strategies and collaborative techniques for task offloading is presented. Computational optimization strategies are classified according to the solutions provided for task offloading via various methods such as algorithmic techniques and Deep Reinforcement Learning (DRL) techniques. Collaborative techniques such as caching, Software Defined Networks (SDN), and Unmanned Aerial Vehicles (UAV) along with a vehicular network for task offloading are extensively reviewed. The security aspect of vehicular task offloading is discussed as well. Furthermore, open issues and future directions of vehicular task offloading are highlighted.

Keywords: IoV, vehicular edge computing, task offloading, security, 6G, multiaccess edge computing, VANET

1.1 Introduction

In the past two decades, there has been a noticeable trend towards the development of intelligent vehicles with substantial developments in communication and computing technologies [1]. It is estimated that the automobile sector will provide the biggest market opportunity for 5G Internet of Things (IoT) solutions by 2023 with the development of intelligent vehicles [2]. IoV, a typical IoT technology application in the Intelligent Transportation System (ITS), is a widely distributed system for information exchange and wireless communication that intelligently supports traffic management, dynamic information services, and vehicle control [3]. Vehicular Technology and Road Side Units (RSU) have progressed rapidly comprising computing units and storage capacities. Utilizing 6G technology, the IoV can achieve seamless connectivity through Space Air Ground Integrated Networks (SAGIN), enabling interoperability between terrestrial and non-terrestrial networks and providing ubiquitous coverage.

With the advent of IoV and 6G, vital developments have emerged in vehicular applications by providing global coverage. The vehicular applications include image-aided navigation, online games, intelligent vehicle control, and other social media applications. Those applications are used to improve traffic efficiency, enhance road safety, as well as provide convenient and comfortable user services [4]. Each of these applications requires ultra-low latency, massive connectivity, high mobility, and scalability support that can be provided by Beyond 5G and Next Generation Networks [5, 6]. Incorporating various applications such as advanced driver assistance systems in smart vehicles poses a significant challenge for in-vehicle computing systems, as they grapple with the escalating demand for processing power. Due to space and power constraints, integrating a supercomputer directly into vehicles is impractical. The limited computing resources, including CPU, memory, and storage, may prove inadequate to meet the rising computational requirements. Thus, it necessitates offloading [7].

Initially, Cloud computing was proposed as an effective solution for resource-constrained vehicles to offload tasks to geographically centralized data centers which improves computation performance and resource utilization [8]. However, the cloud computing architecture makes it very hard to satisfy the real-time processing demands of emerging vehicular applications due to the long propagation delay [9]. Hence, to extend the processing capacity of cloud computing to the edge of a network near the vehicles, Multi-Access Edge Computing (MEC) [10] and Fog Computing have been introduced [11]. Vehicular Edge Computing (VEC) provides processing and storage resources close to vehicular users by integrating MEC and vehicular networks [12]. Figure 1.1 represents the VEC architecture consisting of smart vehicles and infrastructures. Onboard units with resource capabilities are incorporated in smart vehicles that permit close-range wireless transmission i.e., communicate with each other and an RSU. For network accessibility, RSUs are often dispersed along the roads and connected to the backbone network [13]. The illustration of task offloading is depicted in Figure 1.2. Now vehicles may transfer latency-sensitive and computationally heavy tasks to neighboring MEC servers with little overhead. It can significantly alleviate the overload of resource-constrained vehicles. Cellular Base Stations (BS) and RSUs, as well as both, are suitable for placing MEC servers near the network’s edge [14]. Using MEC services with the aid of 6G technology will improve the Quality of Experience (QoE) for vehicular applications. Yet, because of the peculiar features of vehicle networks, particularly the rapid mobility of nodes and the fluctuating channel conditions, it is quite challenging to create an effective edge-enabled task offloading strategy [15].

The MEC servers on distinct BSs may offer a range of services, and the workload on the MEC servers varies over time. MEC servers are less resourceful than cloud servers. To effectively utilize the MEC servers’ resources, consistent load distribution across the MEC servers has to be guaranteed while offloading the tasks from vehicles. The storage and computing capabilities of edge devices are typically constrained. To ensure effective resource usage of the MEC server, some tasks may still need to be performed either locally or on the cloud platform depending on their QoS. Offloading, therefore, requires absolute cooperation between the cloud and the edge. The tasks can also differ in terms of processing overhead, advance, urgency, and other related factors depending on the necessary QoE criteria. As a result, the issue of selecting the best task-offloading strategy for achieving the best performance of an application while effectively using MEC resources arises [16]. It is conceivable to employ both algorithm-based and DRL-based strategies to address this multi-objective optimization problem.

Figure 1.1 Vehicular network.

Figure 1.2 Task offloading.

1.1.1 Study of Existing Surveys

The existing survey papers [1, 17–20] were oriented by vehicular task offloading. Ahmed et al. [17] highlighted the classification of vehicular task offloading based on V2V, V2I, and V2X communication models. Liu et al. [1] presented the classification of vehicular task offloading based on DRL methods as value-based and policy-based solutions leveraging MEC servers, nearby vehicles, and both as edge clouds.

Hamdi et al. [18] majorly analyzed task offloading in vehicular fog computing and elaborated on the fog node selection for task offloading. Boukerche and Soto [19] categorized each step involved in the task offloading process i.e., partitioning, scheduling, and data retrieval as well as analyzed various methods used for these processes. Nevertheless, this survey classifies based on the optimization strategies of task offloading and incorporates associated techniques that enhance vehicular task offloading.

Table 1.1 Existing surveys in vehicular task offloading.

Paper	Year	Categorization criteria	Collaborative techniques	Security

Ahmed et al.

2022

Vehicular Communication modes

No

No

Hamdi et al.

2022

Fog node selection

No

No

Boukerche and Soto

2020

Task offloading process

No

No

Liu et al.

2022

RL/DRL algorithm

No

No

Dziyauddin et al.

2021

Optimization objective (QoS, Energy, Revenue)

Caching

Yes

Our survey

2023

Solution of Optimization Strategies for task offloading

Caching, SDN, UAV

Yes

Although Dziyauddin et al. [20] have also discussed content caching and security along with computational offloading, there exists a research gap in categorizing optimization strategies applied to task offloading problem and collaborative techniques to improve its performance. Table 1.1 summarizes the existing surveys in vehicular task offloading. This survey systematically categorizes various optimization strategies that are used to address the task offloading problem and other techniques associated with...

Produkt-Details

• Erscheint lt. Verlag: 12.3.2025
• Sprache: englisch
• Themenwelt: Mathematik / Informatik ► Informatik ► Theorie / Studium
• ISBN-10: 1-394-31438-8 / 1394314388
• ISBN-13: 978-1-394-31438-6 / 9781394314386