With the advance of artificial intelligence (AI), the Internet of Things (IoT), and 5G communication technologies, various kinds of traffic data from diverse devices can be acquired nowadays, and they can help us look into intelligent transportation systems (ITSs) with a new eye. Graph-based machine learning holds out the potential as a powerful tool for modeling complex structural data relationships and also mining both useful information and temporal patterns which could be used for building powerful analytics for ITS construction. Considering the benefit of graph-based machine learning for ITS, some graph-based machine learning methods/architectures have been proposed. Even though these methods have achieved certain success, there exist various scientific and engineering challenges.

Special Section on AI-Empowered Internet of Things for Smart CitiesOne of the key enablers for smart cities is the Internet of Things (IoT), which can exploit stateof-the-art communication technologies to support advanced services.However, IoT devices and management systems are typically manufactured by multiple vendors with multiple processes and standards.Furthermore, these devices will generate large amounts of data from different sources and types of sensors, which cannot be effectively processed by traditional methods.In addition, the unstructured data in IoT plays an important role for building smart cities, whereas transmitting and processing these unstructured data consume substantial energy.Therefore, the data transmission and processing in IoT for smart cities should be performed in a more intelligent manner.Recently, artificial intelligence (AI) has emerged as a powerful weapon that supports very efficient data analysis and makes accurate decisions on service provisions of various kinds.Combining IoT with advanced AI technology can make the city smarter.AI-empowered solutions, such as deep learning and reinforcement learning, can better process the vast amounts of real-time data that stream from IoT devices to support intelligent services for smart cities.In light of this potential, this special section provides a venue to comprehensively cover algorithms, frameworks, technologies, and applications of AI-empowered IoT for smart cities.After a strict peer review, 10 papers were selected for publication in this special issue.Details of these selected papers are as follows."Power Side-Channel Analysis of RNS GLV ECC Using Machine and Deep Learning Algorithms" by Mehrabi et al. proposes an RNS (Residue Number system) GLV (Gallant Lambert Vanstone) elliptic curve cryptography core that is immune to machine-learning-and deep-learning-based sidechannel attacks.The experimental analysis confirms that the proposed crypto core does not leak any information about the private key, and therefore it is suitable for hardware implementations."ISDNet: AI-Enabled Instance Segmentation of Aerial Scenes for Smart Cities" by Garg et al. proposes ISDNet (Instance Segmentation and Detection Network), a novel network to perform instance segmentation and object detection on visual data captured by UAVs.This work enables aerial image analytics for various needs in a smart city.ISDNet makes use of effective anchors to accommodate varying object scales and sizes.The proposed method obtains state-of-the-art results in the aerial context.In "Robust Facial Image Super-Resolution by Kernel Locality-Constrained Coupled-Layer Regression," Gao et al. design a robust context-patch facial image super-resolution scheme via a KLC2LR (Kernel Locality-constrained Coupled-layer Regression) scheme to obtain the desired high-resolution version from the acquired low-resolution image.The compared experiments in the noisy and noiseless cases have verified that the suggested methodology performs better than many existing predominant facial image super-resolution methods.

The ever-increasing growth of urban populations coupled with recent mobile data usage trends has led to an unprecedented increase in wireless devices, services and applications, with varying quality of service needs in terms of latency, data rate, and connectivity. To cope with these rising demands and challenges, next-generation wireless networks have resorted to cloud radio access network (Cloud-RAN) technology as a way of reducing latency and network traffic. A concrete example of this is New York City's LinkNYC network infrastructure, which replaces the city's payphones with kiosk-like structures, called Links, to provide fast and free public Wi-Fi access to city users. When enabled with data storage capability, these Links can, for example, play the role of edge cloud devices to allow in-network content caching so that access latency and network traffic are reduced. In this paper, we propose HybridCache, a hybrid proactive and reactive in-network caching scheme that reduces content access latency and network traffic congestion substantially. It does so by first grouping edge cloud devices in clusters to minimize intra-cluster content access latency and then enabling cooperative-proactively and reactively-caching using LSTM-based prediction to minimize in-network content redundancy. Using the LinkNYC network as the backbone infrastructure for evaluation, we show that HybridCache reduces the number of hops that content needs to traverse and increases cache hit rates, thereby reducing both network traffic and content access latency.

We propose a blockchain-based, distributed protocol for enabling the deployment of Internet-of-Things (IoT) networks on-demand on top of IoT devices. Specifically, the proposed protocol leverages blockchain technology to: 1) enable distributed and secure authentication, registration, and management of participatory IoT devices; 2) provide fast discovery of IoT resources and scalable and secure instantiation of IoT networks-on-demand; and 3) manage payment operations and ensure reliable fund transfers among the network entities. The proposed protocol relies on a peer-to-peer network communication infrastructure to allow communications among the IoT devices in a distributed manner and uses a self-recovery/self-healing mechanism to ensure robustness against device failure and maliciousness. The protocol also introduces and uses a reputation system to monitor registered devices to keep track of their service delivery quality so that their service delivery reputations could be leveraged for future device selection and mapping. We implemented and evaluated the proposed protocol intensively using simulations and showed that it scales well with network parameters, is resilient to faulty devices, and is robust to 51% attack.

This paper proposes a distributed resource sharing protocol for enabling dynamic deployment of IoT networks on-demand in smart cities. The proposed protocol leverages Blockchain technology to: (i) enable distributed and secure management of IoT devices; (ii) provide fast discovery of IoT resources and scalable on-demand networks; and (iii) ensure reliable fund transfers for service payment among the network entities. The protocol relies on a peer-to-peer network infrastructure to allow communication among the IoT devices in a distributed manner, and uses a self-recovery/self- healing mechanism to ensure robustness against device failure and maliciousness. The protocol also introduces and uses a reputation system to monitor and keep track of services delivered by registered devices for quality of service delivery assurance. We implemented and evaluated the proposed protocol intensively using simulations to assess its effectiveness in terms of scalability to network sizes and robustness to device failures.

This paper proposes efficient strategies that shave Data Centers (DCs)' monthly peak power demand with the aim of reducing the DCs' monthly expenses. Specifically, the proposed strategies allow to decide: i) when and how much of the DC's workload should be delayed given that the workload is made up of multiple classes where each class has a certain delay tolerance and delay cost, and ii) when and how much energy should be charged/discharged into DCs' batteries. We first consider the case where the DC's power demands throughout the whole billing cycle are known and present an optimal peak shaving control strategy for it. We then relax this assumption and propose an efficient control strategy for the case when (accurate/noisy) predictions of the DC's power demands are only known for short durations in the future. Several comparative studies based on real traces from a Google DC are conducted in order to validate the proposed techniques.

We propose an integrated, energy-efficient, resource allocation framework for overcommitted clouds. The framework makes great energy savings by 1) minimizing Physical Machine (PM) overload occurrences via VM resource usage monitoring and prediction, and 2) reducing the number of active PMs via efficient VM migration and placement. Using real Google data consisting of a 29-day traces collected from a cluster containing more than 12K PMs, we show that our proposed framework outperforms existing overload avoidance techniques and prior VM migration strategies by reducing the number of unpredicted overloads, minimizing migration overhead, increasing resource utilization, and reducing cloud energy consumption.

Over the past couple of decades, transportation systems have begun to receive widespread attention from the scientific community and emerged toward Intelligent Transportation Systems (ITS). Effective vehicular connectivity techniques can significantly enhance efficiency of travel, reduce traffic incidents and improve safety, and alleviate the impact of congestion; devising the ITS experience. Furthermore, during the past decades, the volume and density of vehicles increased significantly, especially the road traffic; this lead to a dramatic increase in the number of accidents and congestion, with negative impacts on the economy, environment, and quality of people’s lives. In particular, according to the World Health Organization (WHO), road traffic injuries are estimated to be the leading cause of death for young people aged 15–29 and the ninth cause of death worldwide in 2015. The enabling communication technologies are intended to realize the frameworks that will spur an array of applications and use cases in the domain of road safety, traffic efficiency, and driver’s assistance. Although these applications will allow the dissemination and gathering of useful information among vehicles and between transportation infrastructure and vehicles in pursuance of assisting drivers to travel safely and comfortably, much effort is required to implement these practices for the success of these applications.

Large urban communication networks such as smart cities are an ecosystem of devices and services cooperating to address multiple issues that greatly benefit end users, cities, and the environment. LinkNYC is a first-of-its-kind urban communications network aiming to replace all payphones in the five boroughs of New York City (NYC) with kiosk-like structures providing free public Wi-Fi. We consolidate these networks with standalone edge cloud devices known as cloudlets and introduce geographically distributed content delivery cloudlets (CDCs) to store popular Internet content closer to end users; essential in environments with diverse and dynamic content interests. A content-centric and delivery framework is proposed leveraging NYC's population densities and CDCs for interest-based in-network caching. Analysis shows that although the adoption of multiple CDCs dramatically improves overall network performance, advanced caching policies are needed when considering increased content heterogeneity. Thus, we propose popularity-driven and cooperation-based caching policies at individual CDCs to account for user and content dynamics over time. The amalgamation of urban population densities, multiple CDC placements and smarter caching techniques helps exploit the ultimate benefits of a content-centric urban communications network and dramatically improves overall network performance and responsiveness. Our proposed solutions are validated using LinkNYC as a use-case.

Internet of Things (IoT) applications can benefit greatly from cloud-hosted message broker services that utilize publish-subscribe communications. The operators of IoT cloud-hosted services are often interested in delivering services that maximize their revenue given quality of service guarantees. In this paper, we formulate the problem of maximizing the profit of the service provider given the prior knowledge of the request sequence as an integer linear program and prove that it is strongly NP-complete, and thus there is no fully polynomialtime approximation scheme for the problem, unless P = NP. Due to the above-mentioned problem and the difficulty of obtaining the request sequence in advance in real-world scenarios, we propose an auction-based online algorithm that does not require the prior knowledge of the request sequence. We prove that the competitive ratio of the online algorithm is (9(log(N)), where N is the number of cloud zones that host the publish- subscribe services. Moreover, we show that no online algorithm can achieve a competitive ratio better than Ω(log(N)). Therefore, our online algorithm achieves the optimal competitive ratio in the asymptotic sense. Our simulations, based on real data traces, show that our algorithm achieves up to 83% more profit compared to a heuristic approach, while consuming 60% less resources.

One of the greatest challenges of an enterprise's service center is to ensure that their engineers and customers are provided with the right information in a timely fashion. For this purpose, modern organizations operate a wide range of information support systems to assist customers with critical service requests and to provide proactive monitoring, where possible, to prevent service requests from occurring in the first place. It is often the case that relevant information is scattered over the Internet and/or maintained on disparate systems, buried in large amount of noisy data, and in heterogeneous formats, thereby complicating the access to reusable knowledge and extending the response time to reach a resolution. To address these challenges, in this paper we propose an effective knowledge mining solution to improve the quality of service request resolution. We model the service resolution problem as an online search and classification problem, and use domain knowledge in the form of ontology to guide effective machine learning. Our proposed solution has been extensively evaluated with experiments and has been used in a real enterprise customer center.

We propose efficient control strategies for deciding the amount of energy that a battery needs to charge/discharge over time with the objective of minimizing the Peak Charge and the Energy Charge components of the Data Center (DC) electricity bill. We consider first the case where the DC's power demands throughout the whole billing cycle are known and we present an optimal peak shaving control strategy for a battery that has certain leakage and conversion losses. We then relax this assumption and propose an efficient battery control strategy when we only know predictions of the DC's power demands in a short duration in the future. Several comparative studies are conducted based on real traces from a Google DC in order to validate the proposed techniques.

Managing cloud resources in a way that reduces the consumed energy while also meeting clients demands is a challenging task. In this paper, we propose an energy-aware resource allocation framework that: i) places the submitted tasks (elastic/inelastic) in an energy-efficient way, ii) decides initially how much resources should be assigned to the elastic tasks, and iii) tunes periodically the allocated resources for the currently hosted elastic tasks. This is all done with the aim of reducing the number of ON servers and the time for which servers need to be kept ON allowing them to be turned to sleep early to save energy while meeting all clients demands. Comparative studies conducted on Google traces show the effectiveness of our framework in terms of energy savings and utilization gains.

Energy efficiency has recently become a major issue in large data centers due to financial and environmental concerns. This paper proposes an integrated energy-aware resource provisioning framework for cloud data centers. The proposed framework: i) predicts the number of virtual machine (VM) requests, to be arriving at cloud data centers in the near future, along with the amount of CPU and memory resources associated with each of these requests, ii) provides accurate estimations of the number of physical machines (PMs) that cloud data centers need in order to serve their clients, and iii) reduces energy consumption of cloud data centers by putting to sleep unneeded PMs. Our framework is evaluated using real Google traces collected over a 29-day period from a Google cluster containing over 12,500 PMs. These evaluations show that our proposed energy-aware resource provisioning framework makes substantial energy savings.

We propose a credit-based resource allocation technique for dynamic spectrum access that is robust against malicious and selfish behaviors and ensures good overall system fairness performance while also allowing spectrum users to achieve high amounts of service. We also propose a new objective function that, when combined with the proposed credit-based technique, leads to further improvements of the system fairness performance. Our proposed techniques overcome user misbehavior by masking the impact of the users' pursued private objectives on the overall system performance. They also improve fairness among users by allocating service to users adaptively by accounting for how much service each user has received in the past. Our simulation results show that our proposed techniques maintain high system performance by allowing users to achieve high amounts of service and by ensuring fair allocation of spectrum resources among users even in the presence of misbehaved users. Using simulations, we also show that these high performances are also achievable under various different network scenarios.

We propose a credit-based resource allocation technique for wireless systems with dynamic spectrum access (DSA) capability, such as multichannel wireless sensor networks. The proposed technique is robust against selfish and malicious behaviors by achieving high performance independently of what users choose to pursue as their objectives. It also improves fairness by ensuring that different users are allocated equal amounts of spectrum service. Using simulations, we show that the proposed credit-based technique allows users to achieve high rewards/service even in the presence of misbehaved users that choose to pursue their greedy and selfish goals. We also show that it reduces the standard deviation of users' amounts of received service drastically, and hence, improves fairness among users substantially.

The recent emergence and success of cloud-based services has empowered remote sensing and made it very possible. Cloud-assisted remote sensing (CARS) enables distributed sensory data collection, global resource and data sharing, remote and real-time data access, elastic resource provisioning and scaling, and pay-as-you-go pricing models. CARS has great potentials for enabling the so-called Internet of Everything (IoE), thereby promoting smart cloud services. In this paper, we survey CARS. First, we describe its benefits and capabilities through real-world applications. Second, we present a multilayer architecture of CARS by describing each layer's functionalities and responsibilities, as well as its interactions and interfaces with its upper and lower layers. Third, we discuss the sensing services models offered by CARS. Fourth, we discuss some popular commercial cloud platforms that have already been developed and deployed in recent years. Finally, we present and discuss major design requirements and challenges of CARS.

We develop efficient coordination techniques that support inelastic traffic in large-scale distributed dynamic spectrum access (DSA) networks. By means of any learning algorithm, the proposed techniques enable DSA users to locate and exploit spectrum opportunities effectively, thereby increasing their achieved throughput (or “rewards” to be more general). Basically, learning algorithms allow DSA users to learn by interacting with the environment, and use their acquired knowledge to select the proper actions that maximize their own objectives, thereby “hopefully” maximizing their long-term cumulative received reward. However, when DSA users' objectives are not carefully coordinated, learning algorithms can lead to poor overall system performance, resulting in lesser per-user average achieved rewards. In this paper, we derive efficient objective functions that DSA users can aim to maximize, and that by doing so, users' collective behavior also leads to good overall system performance, thus maximizing each user's long-term cumulative received rewards. We show that the proposed techniques are: (i) efficient by enabling users to achieve high rewards, (ii) scalable by performing well in systems with a small as well as a large number of users, (iii) learnable by allowing users to reach up high rewards very quickly, and (iv) distributive by being implementable in a decentralized manner.

We develop objective functions for large-scale distributed dynamic spectrum access (DSA) networks that, by means of any learning algorithm, enable DSA users to locate and exploit spectrum opportunities effectively, thereby increasing their achieved throughput (or "rewards" to be more general). We show that the proposed functions are: (i) optimal by enabling users to achieve high rewards, (ii) scalable by performing well in systems with a small as well as a large number of users, (iii) learnable by allowing users to reach up high rewards very quickly, and (iv) distributed by being implementable in a decentralized manner.

In this paper, we present a novel algorithm that extracts frequent anomalous behaviors based on multi-scale trend analysis of individual network parameters. The proposed Frequent Anomalous Behavior Mining (FABM) algorithm utilizes multiple levels of time-scale analysis to reveal the frequent anomalous behaviors. This makes the proposed algorithm robust to unreliable, redundant, incomplete and contradictory information. FABM is simple, has low order polynomial computational complexity of O(n2), the patterns identified by FABM require space complexity of O(n) to be stored in the knowledge base of the prediction engine, provides quick and accurate response and can be easily adapted to a distributed environment. Moreover, the empirical results gathered show that using FABM an efficient prediction engine can be realized with high true positive and true negative rates.

In this paper we describe an effective new technique by which to guarantee cooperativeness in Hybrid Radio-Frequency/Free Space Optics (RF/FSO) networks. Our approach is based on a novel Bayesian game-theoretic model, and uses a pricing scheme in which each destination node pays some amount of virtual money to the source node in order to acquire a reliable connection. We describe both single-stage and multi-stage solutions for the game in terms of its Nash and Perfect Bayesian Equilibriums. Pure strategies are found when the required conditions are met; otherwise the game is played as a mixed-strategy. Our numerical results quantify the inherent tradeoffs involved in changing the game's parameters vis-a-vis the equilibrium player strategies and game's outcomes.

The majority of vehicle-to-vehicle (V2V) protocols presented in the literature utilize a combination of controlled access techniques (e.g., TDMA, FDMA, CDMA) to provide prioritized access to the communication medium. While these protocols split the capacity of the communication link to support a maximum number of users with different priority levels, our work utilizes a combination of CDMA and TDMA techniques to increase the number of concurrent users sharing the bandwidth. The proposed protocol provides an adaptive scheme that classifies the vehicle's messages as "urgent" or "non-urgent" based on its statistical parameters (e.g., speed, acceleration, directional stability) and the state of its neighboring vehicles (e.g., normal or urgent) . Furthermore, the proposed protocol achieves intelligent scheduling and allocation of messages and the underlying bandwidth to minimize the end-to-end communication delay and the costs associated with the deployment of vehicle infrastructure integration (VII) without the need for road-side equipment.

Intelligent service management techniques play an important role in the continuously and rapidly evolving area of technologically advanced services. High tech companies are looking for better ways to deliver and preserve services to their customers in a competitive way. This paper introduces a new architecture for a scalable service monitoring and support system called Call Home Analysis and Response System (CHARS). The proposed system utilizes data de-noising and filtering techniques to meet the service management requirements of large-scale service deployments. The system utilizes intra and inter-element correlation of events to enhance the services delivered to end- users. Our results demonstrate that the proposed system is effective in covering the performance and fault management aspects for large-scale deployments of advanced services.

Mission-oriented MANETs are characterized by implicit common group objectives which make inter-node cooperation both logical and feasible. We propose new techniques to leverage two optimizations for cognitive radio networks that are specific to such contexts: opportunistic channel selection and cooperative mobility. We present a new formal model for MANETs consisting of cognitive radio capable nodes that are willing to be moved (at a cost). We develop an effective decentralized algorithm for mobility planning, and powerful new Altering and fuzzy based techniques for both channel estimation and channel selection. Our experiments are compelling and demonstrate that the communications infrastructure-specifically, connection bit error rates-can be significantly improved by leveraging our proposed techniques. In addition, we find that these cooperative/opportunistic optimization spaces do not trade-off significantly with one another, and thus can be used simultaneously to build superior hybrid schemes. Our results have significant applications in high-performance mission-oriented MANETs, such as battlefield communications and domestic response & rescue missions.

In this paper we present new energy-efficient techniques to lower the packet-level error rates of application-layer connections in wireless ad-hoc networks. In our scheme, each application-layer connection is implemented at the physical level by an overlay network. Data packets submitted at the connection source are checksummed and replicated, flowing breadth-first across the overlay network towards the destination. The destination delivers the first error-free copy of each packet, in order, to the application layer, dropping packets that are corrupt or duplicate. Specifically in this paper, we consider overlays consisting of multiple parallel node-disjoint multi-hop paths. We compare this overlay scheme with the traditional scheme in which the source transmits to the destination along a single minimum-hop path. We show that even when the two schemes are constrained by identical power consumption bounds, an overlay scheme that uses multiple multi-hop paths provides significantly lower packet-level error rates in many common situations. We describe the relationship between packet error rate, the number of paths, and the lengths of each path, and show that the qualitative nature of the relationship changes significantly, depending on available power budget.

This paper presents a service on demand management framework to support dynamic service creation, and a system which instantiates this framework for an Ethernet-To-The-Home/Business (ETTx) service environment. The service-on-demand framework delivers two major functions: (1) it allows service providers to create new types of service dynamically through a policy-driven "service building" process; (2) it allows end users to subscribe to a service and have their service orders fulfilled in a fully automated, near real time, and "zero touch" fashion. We have applied this framework to build an industrial strength service fulfillment system geared towards ETTx access networks. Those networks represent one of the most dynamic service creation environments today.

The wireless market can be divided into mobile/cellular and fixed wireless. The mobile/cellular world is progressively transitioning from 2G (e.g. GSM, 2G-CDMA) to 2.5G (e.g. GPRS, 2.G-CDMA) to 3G (e.g. CDMA-2000, UMTS) technologies. The broadband fixed wireless technology is targeted as one of the "last mile" services available to residents and businesses. It typically requires a licensed frequency with a specific frequency range. A typical broadband fixed wireless system consists of a hub system servicing multiple subscriber end systems over a wireless radio frequency (RF). Examples of fixed wireless solutions include the Multi-channel Multi-point Distribution System (MMDS) and the Local Multi-point Distribution System (LMDS). This paper defines the OSS architecture, based on the Telecom Operation Map framework of the TeleManagement Forum, for mobile and fixed wireless networks. This include the FCAPS five functional areas: fault configuration management, accounting performance management, and management, management, management as well as service restoration, engineering, data collection, service activation, and network planning. The proposed architecture will give service providers the capability to manage their next generation networks in a way that gives them FCAPS in a way that best suits the technology and their business processes.

The intensity of exogenous traffic is a critical input to a variety of network functions, including routing, flow control, and capacity planning. Currently, however, the scope of measurements in a large scale, circuit-switched network is quite limited and does not include gathering of the exogenous traffic intensity in real-time. This paper formulates methods of estimating the exogenous traffic intensity, without requiring additional measurements beyond what is currently available. The performance of the proposed methods is compared based on network simulations and is shown to be accurate and robust.>