The advent of microservice architecture enables complex cloud applications to be realized via a set of individually isolated components, increasing their flexibility and performance. As these applications require massive computing resources, graphics processing units (GPUs) are being widely used as high-speed parallel computing devices to meet the stringent demands. Although current GPUs allow application components to be executed concurrently via spatial multitasking, they face several challenges. The first challenge is allocating the computing resources to components dynamically to maximize efficiency. The second challenge is avoiding performance degradation caused by the data transfer overhead between the components. To address these challenges, we propose an efficient GPU resource management technique that dynamically allocates GPU resources to application components. The proposed method allocates resources based on component workloads and uses online performance monitoring to guarantee the application's performance. We also propose a GPU memory manager to reduce the data transfer overhead between components via shared memory. Our evaluation results indicate that the proposed dynamic resource allocation method improves application throughput by up to 134.12% compared to the state-of-the-art spatial multitasking techniques. We also show that using a shared memory results in 6x throughput improvement compared to the baseline User Datagram Protocol (UDP)-based technique.

Applications deployed in clouds are susceptible to performance degradation due to diverse underlying causes such as infrastructure faults. To maintain the expected availability of these applications, Machine Learning (ML) models can be used to predict the impending application performance degradations to take preventive measures. However, the prediction accuracy of these ML models, which is a key indicator of their performance, is influenced by several factors, including training data size, data sampling intervals, input window and prediction horizon. To optimize these data-related parameters, in this paper, we propose a surrogate-assisted multi-objective optimization algorithm with the objective to maximize prediction model accuracy while minimizing the resources consumed for data collection and storage. We evaluated the proposed algorithm through two use cases focusing on the prediction of Key Performance Indicators (KPIs) for a 5 G core network and a web application deployed in two Kubernetes-based cloud testbeds. It is demonstrated that the proposed algorithm can achieve a normalized hypervolume of 99.5% relative to the optimal Pareto front and reduce search time for the optimal solution by 0.6 hours compared to other surrogates and by 3.58 hours compared to using no surrogates.

To implement a Network Service (NS) within a Network Function Virtualization (NFV) environment, it is essential to create a sequence of connected Virtual Network Functions (VNFs), known as a VNF Forwarding Graph (VNF-FG), and then embed it onto the substrate network. The emergence of VNF decomposition as a new functional architecture allows VNFs to be broken down into smaller sub-functions, offering enhanced flexibility, resource sharing, and scalability. VNF decomposition can significantly reduce VNF embedding costs since different sub-functions can be efficiently reused by multiple network requests. However, when VNFs are decomposed into multiple sub-functions, selecting the appropriate decomposition option for each VNF and constructing the VNF-FG to embed onto the substrate network poses a significant challenge in NFV resource allocation (NFV-RA). A key challenge is identifying the optimal decomposition option among all possible choices for VNF embedding. In this paper, we introduce a cost-aware algorithm designed to address the topological decomposition of VNF-FGs, focusing on minimizing embedding costs while meeting specified service requirements. We formulate the VNF topology decomposition problem using Integer Linear Programming (ILP) to select the best decomposition option and minimize the embedding cost. Furthermore, we propose four efficient heuristics for different topologies to identify the optimal decomposition options for network embedding. Simulation results demonstrate that our proposed algorithm outperforms existing benchmarks in terms of embedding costs and achieves execution times that are up to 95% better than the SE approach.

During a Remote Robotic Surgery (RRS) session, multimodal data traffic with different requirements is initiated. In order to achieve a cost-effective deployment of such a system, it is crucial to tailor resource allocation policies based on the different quality of service (QoS) requirements of each data traffic. In this paper, we focus on resource allocation in a 5G-enabled tactile Internet RRS system using network function virtualization (NFV). In particular, we investigate the joint placement and scheduling of Virtualized Network Functions (VNFs) in a RRS system under both deterministic and dynamic settings. An integer linear program (ILP) is used to formulate the problem. Due to its high computational complexity, we first propose an efficient greedy algorithm to solve the ILP under deterministic settings. Simulation results show that our proposed algorithm achieves near-optimal performance and outperforms the benchmark solutions in terms of cost and admission rate. It can reduce cost by up to 37% and improve admission rate by up to 34% while satisfying both latency and reliability constraints. Furthermore, our results show that modeling the multimodal data traffic by multiple VNF Forwarding Graphs (VNF-FGs) with different QoS requirements achieves a significant gain in terms of cost and acceptance rate compared to modeling it by a single VNF-FG with the most stringent requirements. We then considered a dynamic environment where latency variations and traffic arrivals may occur over time. Using the principles of optimal stopping theory, we propose an adaptive dynamic scheduler that is capable of triggering recalculations of the existing optimal solution based on the observed cumulative number of traffic arrivals and latency violations without the need for predictions. Our proposed optimal scheduler minimizes the migration cost compared to other schedulers.

Applications deployed in edge cloud environments can have stringent requirements such as high throughput and high availability. However, these applications may suffer from performance degradation caused by various underlying reasons such as infrastructure-related faults. Handling application performance degradation proactively is thus critical for maintaining the application Quality-of-Service (QoS). This can be achieved through predicting application performance degradation using Machine Learning (ML) models. The performance of these ML models may degrade over time due to changes in the relevancy of features used for training the ML model for application performance degradation, i.e., feature drift. In this paper, we predict application performance degradation in edge clouds and propose a framework for adapting to the feature drifts that may occur in this environment. This framework detects a feature drift using performance of the prediction model as well as feature importance, and updates the features and adapts the prediction model to the drift considering the severity of the feature drift. We have built a proof-of-concept of our proposed framework on a Kubernetes testbed. It is demonstrated that the proposed framework can achieve up to 9.1% higher F1-score compared to Dynamic Correlation-based Feature Selection (DCFS) approach for feature drift adaptation from the literature.

Intent-Based Networking (IBN) has recently gained interest to realize intelligent cloud management systems. IBN makes cloud consumption user-friendly by allowing users to specify “what needs to be done” rather than “how it should be done”. In this paper, we address the problem of service graphs generation in IBN, while considering functional requirements derived from user intent. We propose a solution that uses a domain ontology to translate the intents into their functional requirements which are specified in terms of initial services. Then, we use a service catalog along with the initial services to generate all possible service graphs that can meet the functional requirements of the given intent. Our evaluations show that the proposed solution achieves up to 84% improvement of the execution time compared to the existing benchmarks.

Network Function Virtualization (NFV) is a network architecture that separates network functions from dedicated hardware, implementing them as software modules known as Virtual Network Functions (VNFs), which are executed in virtual machines or containers. NFV increases the deployment flexibility and agility within operator networks and reduces the operating and capital expenditures significantly. In NFV, migration of VNFs can significantly reduce the embedding cost. However, stringent latency requirements between VNFs can make them tightly coupled, thus hindering each VNF from being migrated individually, and resulting in poor performance. One of the main challenges in an NFV environment is therefore to migrate a cluster of VNFs to minimize the embedding cost. In this paper, we aim to solve the problem of cluster VNF migration by considering the given inter-VNF latency requirements. We formulate the VNF migration problem as an Integer Linear Programming (ILP) and present two scalable and efficient algorithms for migrating a cluster of VNFs. Through extensive experiments, we show that our proposed algorithms are highly effective. They reduce the total embedding cost by 14% compared to the existing heuristics, while being much more scalable in terms of execution time compared to the brute-force approach.

Ensuring the quality of service of applications deployed in inherently complex and fault-prone cloud environments is of utmost concern. While machine learning based fault management solutions help attain the desired reliability, they require labeled cloud metrics data for training and evaluation. Furthermore, high dynamicity of cloud environments brings forth emerging data distributions, which necessitate frequent labeling of data for model adaptation. We propose a test suite-based active learning framework for automated labeling of cloud metrics data with the corresponding cloud system state while accounting for emerging fault patterns and data or concept drifts. We have implemented our solution on a cloud testbed and introduced various emerging data distribution scenarios to evaluate the proposed framework’s labeling efficacy over known and emerging data distributions. According to our results, the proposed framework achieves about 41% higher weighted F1-score and 34% higher average Area Under the One-vs-Rest Receiver Operating Characteristic Curve (AUC) score than a system without any adaptation for emerging data distributions.

To ensure reliability and service availability, next-generation networks are expected to rely on automated anomaly detection systems powered by advanced machine learning methods with the capability of handling multi-dimensional data. Such multi-dimensional, heterogeneous data occurs mostly in today’s Industrial Internet of Things (IIoT), where real-time detection of anomalies is critical to prevent impending failures and resolve them in a timely manner. However, existing anomaly detection methods often fall short of effectively coping with the complexity and dynamism of multi-dimensional data streams in IIoT. In this paper, we propose an adaptive method for detecting anomalies in IIoT streaming data utilizing a multi-source prediction model and concept drift adaptation. The proposed anomaly detection algorithm merges a prediction model into a novel drift adaptation method resulting in accurate and efficient anomaly detection that exhibits improved scalability. Our trace-driven evaluations indicate that the proposed method outperforms the state-of-the-art anomaly detection methods by achieving up to an 89.71% accuracy (in terms of Area under the Curve (AUC)) while meeting the given efficiency and scalability requirements.

The growth of streaming data originating from Internet of Things (IoT)-based Industry 4.0 opens doors to real-time analytics of time-sensitive services. However, this ever-increasing amount of data inevitably leads to anomalies, resulting in considerable risks for time-sensitive applications. Thus, real-time detection of anomalies is critical to prevent impending failures and resolve them in time. Given that the problem is to detect application-level anomalies in real time, we develop a deep learning-based technique, which integrates time-series data inference with a Long-Short Term Memory (LSTM)-based prediction model. Our proposed method relies on a novel metric called Sequence Inconsistency Distance (SID), which determines the abnormality likelihood of a target record in real time. Our trace-driven evaluations indicate that the proposed method achieves up to a 92.6% performance gain compared to the current state-of-the-art anomaly detection methods in terms of true positive and false positive rate while meeting the essential efficiency requirements.

Network Function Virtualization (NFV) enables the decoupling of network functions from dedicated hardware to run them as software instances on commodity servers through virtualization, replacing hardware-based network functions with software-based Virtual Network Functions (VNFs). In this paper, we study the joint problem of VNF decomposition and migration to address VNF embedding in NFV resource allocation (NFV-RA). More specifically, we investigate how VNF migration and VNF decomposition can be mutually beneficial to minimize the embedding cost of network services. After presenting a novel formulation of the problem as an integer linear programming (ILP), we validate it by CPLEX and show that our joint VNF decomposition and migration approach can outperform the VNF decomposition-only approach by 20% in terms of embedding cost.

To realize a Network Service (NS) in a Network Function Virtualization (NFV) network, it is needed to form an ordered set of connected Virtual Network Functions (VNFs), commonly referred to as VNF Forwarding Graph (VNF-FG), and then embed it onto the substrate network. Forming a VNF-FG is a challenging step of NFV resource allocation (NFV-RA), especially when the VNFs can be further decomposed into different sub-functions. In this paper, we propose a cost-aware algorithm to solve the problem of topological decomposition of VNF-FGs with the main objective of minimizing the embedding cost while satisfying the given service requirements. The simulation results indicate that our proposed algorithm outperforms the existing benchmark in terms of embedding cost, while being significantly scalable compared to the brute-force approach.

Next-generation networks will enable applications that are expected to be highly reliable, always available, with guaranteed Quality-of-Service (QoS). Distributed, heterogeneous edge clouds are key enablers for these applications. However, applications deployed in such networks may suffer from performance degradation caused by various infrastructure-related faults. Preventing performance degradations by predicting them using Machine Learning (ML)-based analytics and handling them proactively is thus critical for maintaining the application QoS. Predicting performance degradations can be challenging due to the diversity of the underlying causes. In this paper, we propose an automated feature selection system that uses causal-temporal analysis to find the infrastructure metrics that have causal relationships with application metrics. The selected features are further used to train ML models for predicting application performance degradation. We have validated a proof-of-concept of our system on a Kubernetes testbed, where it is demonstrated that the proposed feature selection system is up to 17.9 times faster than Recursive Feature Elimination (RFE). Using the features selected by the proposed system, the ML models could predict performance degradation with up to 12.7% higher F1-score compared to using historical values of application Key Performance Indicator (KPI) to forecast its future values.

Ensuring the quality of service of applications deployed in inherently complex and fault-prone cloud environments is of utmost concern. While machine learning based fault management solutions help attain the desired reliability, they require labeled cloud metrics data for training and evaluation. Furthermore, high dynamicity of cloud environments brings forth emerging data distributions, which necessitate frequent labeling of data for model adaptation. We propose a test suite-based active learning framework for automated labeling of cloud metrics data with the corresponding cloud system state while accounting for emerging fault patterns and data or concept drifts. We have implemented our solution on a cloud testbed and introduced various emerging data distribution scenarios to evaluate the proposed framework’s labeling efficacy over known and emerging data distributions. According to our results, the proposed framework achieves a 41% higher weighted Fl-score and a 34% higher average AUC score than a system without any adaptation for emerging data distributions.

The Industrial Internet of Things (IIoT) applications is usually associated with stringent latency requirements. An anomaly in an IIoT edge network deteriorates the performance and thus needs a real-time Root Cause Analysis (RCA) to identify the anomalous node and provide robust network infrastructure. In this paper, we present an automated, real-time RCA technique to identify the network-level root cause nodes. We use a deep learning-based approach, exploiting a Graph Neural Network (GNN) to identify root cause nodes. In GNN-RCA, we used a sampling technique and optimized aggregator function to reduce detection time. We have shown that the proposed GNNRCA method outperforms the existing benchmarks in terms of classification score and execution time.

By decoupling network functions from the underlying hardware, Network Function Virtualization (NFV) allows application components to be implemented as sets of Virtual Network Functions (VNFs) chained in a specific order, represented by VNF-Forwarding Graphs (VNF-FG). Fog computing is instrumental to tap into the full potential of NFV by deploying VNFs in close proximity to end-users, thus decreasing the latency significantly. However, the mobility of end-users and the fog nodes, and the limited fog nodes coverage results in service discontinuity and may increase application delay. Application component migration offers great potential to address this issue. In this paper, we propose a component migration strategy in an NFV-based hybrid cloud/fog system considering the mobility of both end-users and fog nodes. We use the Gauss-Markov mobility model and a random walk mobility model for fog nodes and end-user devices, respectively. We modeled the problem mathematically, which minimizes the aggregated weighted function of application delay and cost. However, considering the mobility of both end-users and fog nodes makes the problem quite complex. Hence, we propose a Deep Reinforcement Learning (DRL) approach to decide where and when to migrate application components and to achieve rapid decision-making. Simulation results demonstrate that the proposed scheme performs well. It offers favorable convergence and outperforms existing algorithms in terms of application delay and migration costs.

Network Function Virtualization (NFV) is a network service deployment technology that reduces capital and operational costs while yielding flexibility and scalability for service operators. As such, an ordered chain of Virtual Network Functions (VNFs), known as a VNF Forwarding Graph (VNF-FG), should be composed and embedded into the underlying substrate network. In the literature, the composition and embedding stages of VNF-FGs are usually targeted separately, which may result in undesired solutions. In this paper, we propose our joint VNF-FG composition and embedding solution, which considers the variations of service demands while also accounting for dynamic network conditions. Specifically, our proposed solution relies on deep reinforcement learning empowered by two components for estimating dynamic parameters: network resource utilization and service demand analyzers. Moreover, to efficiently explore the problem’s large discrete action space, we utilize a specialized branching Q-network and enhance it with an action filtering mechanism. We evaluated our proposed method against joint and disjoint composition and embedding heuristics as well as versus other deep learning-based methods. Our results show that the proposed method can achieve up to a 95% improvement of embedding cost compared to our benchmarks.

Dynamic and zero-touch management is expected to be the key feature of next-generation 6G networks. Network Function Virtualization (NFV) is one of the key technologies for realizing such management through software-based networks. Despite great benefits offered by NFV, deploying network services (NSs) in NFV ecosystems remains a challenge, especially for latency-sensitive NSs, as they demand stringent latency requirements and fast service provisioning. Specifically, service graphs should be embedded into an infrastructure such that these requirements are satisfied while optimizing network operator’s objectives. To cope with the scalability of optimization-based approaches, heuristic methods are known as promising alternatives to find a satisfactory solution within an acceptable execution time. However, existing VNF embedding heuristics still suffer from the so-called causality issue, which may degrade the embedding solution quality. The causality issue means that embedding decisions cannot be optimally determined before all neighboring dependencies are known. To this end, we introduce our${h}$-horizon sequential look-ahead greedy embedding framework, which provides efficient embedding and re-embedding strategies to alleviate the impact of the causality issue. The simulation results indicate that our proposed algorithm significantly improves embedding cost, compared to the existing heuristic algorithms while being much more scalable than an optimization-based approach.

In traditional Wireless Sensor Networks (WSNs) the applications are embedded in sensor nodes, making them application-oriented, domain-specific devices. Virtualization is a promising approach to allow several applications to use the physical resources of the same deployed WSN. Given that WSN sensor nodes hold limited available energy, efficient allocation and utilization of resources become important. In this paper, we study the problem of virtual network embedding in virtualized WSNs in a static setting. Our objective is to minimize the overall energy consumption by embedding the application virtual network requests onto the physical WSN substrate network. We formulate the problem as an Integer Linear Programming (ILP), accounting for latency as the service-level agreement (SLA). We then propose our so-called E2NE heuristic to achieve a near-optimal solution in a computationally efficient manner. Our results show that embedding virtual sensor networks in a virtualized WSN leads to more energy efficiency than embedding the same networks in traditional WSN without any support for virtualization. This is largely due to the fact that more physical sensors need to be activated in traditional WSN because each physical sensor can run one and only one task at a time.

Remote robotic surgery is one of the most interesting Tactile Internet (TI) applications. It has a huge potential to deliver healthcare services to remote locations. Moreover, it provides better precision and accuracy to diagnose and operate on patients. Remote robotic surgery requires ultra-low latency and ultra-high reliability. The aforementioned stringent requirements do not apply for all the multimodal data traffic (i.e., audio, video, and haptic) triggered during a surgery session. Hence, customizing resource allocation policies according to the different quality-of-service (QoS) requirements is crucial in order to achieve a cost-effective deployment of such system. In this paper, we focus on resource allocation in a softwarized 5G-enabled TI remote robotic surgery system through the use of Network Functions Virtualization (NFV). Specifically, this work is devoted to the joint placement and scheduling of application components in an NFV-based remote robotic surgery system, while considering haptic and video data. The problem is formulated as an integer linear program (ILP). Due to its complexity, we propose a greedy algorithm to solve the developed ILP in a computationally efficient manner. The simulation results show that our proposed algorithm is close to optimal and outperforms the benchmark solutions in terms of cost and admission rate. Furthermore, our results demonstrate that splitting application traffic to multiple VNF-forwarding graphs (VNF-FGs) with different QoS requirements achieves a significant gain in terms of cost and admission rate compared to modeling the whole application traffic with one VNF-FG having the most stringent requirements.

Fault management systems that use real-time analytics based on Machine Learning (ML) help provide the reliability required in edge clouds, though they suffer from frequent changes in data distribution (concept drift) caused by highly dynamic traffic in edge clouds, requiring frequent adaptations of the ML model. We propose an automated concept drift handling framework for fault prediction in edge clouds using Reinforcement Learning (RL) to select the most appropriate drift adaptation method as well as the amount of data needed for adaptation, while considering edge cloud operator’s requirements. We implemented an edge cloud testbed and introduced infrastructure and network faults to it in the presence of abrupt and incremental concept drift. According to the obtained results, our proposed framework achieves up to 40% higher accuracy compared to a system without drift handling, and up to$13\times $and$30\times $less regret for selecting adaptation methods and amount of data, respectively, compared to other approaches.

6G networks are envisioned to be trustworthy systems that can self-maintain their reliability and service availability. The ability to automatically recognize and predict any faults or failures that occur while delivering services is a step towards realizing such systems. Real-time analytics based on Artificial Intelligence (AI)/Machine Learning (ML) techniques provides the required functionality. However, in the highly dynamic and complex environment of 6G services, the distribution of data feeding to the ML models is subject to change over time (i.e., concept drift), which makes persistent model accuracy challenging without any frequent model retraining. To address this issue, in this paper, we propose a system that predicts the faults in an edge cloud environment, where the prediction models are trained and automatically adapted to the concept drifts via Transfer Learning (TL). To present the effectiveness of our system, we implemented an edge cloud testbed and introduced CPU over-utilization and network congestion fault to it in the presence of concept drift. We used Convolutional Neural Network (CNN), Long-Short Term Memory (LSTM), and a combination of them as fault prediction models. For detecting the drifts, we have implemented and compared four drift detection methods, and we utilized three TL scenarios for adapting the model to the drift. Our results indicate the superiority of our system in maintaining a persistent accuracy in the presence of concept drift compared to a prediction system without a drift handling entity.

Remote robotic surgery, one of the most interesting 5G-enabled Tactile Internet applications, requires an ultra-low latency of 1 ms and high reliability of 99.999%. Communication disruptions such as packet loss and delay in remote robotic surgery can prevent messages between the surgeon and patient from arriving within the required deadline. In this paper, we advocate for scalable Gaussian process regression (GPR) to predict the contents of delayed and/or lost messages. Specifically, two kernel versions of the sequential randomized low-rank and sparse matrix factorization method ($\ell _{1}$-SRLSMF and SRLSMF) are proposed to scale GPR and address the issue of delayed and/or lost data in the training dataset. Given that the standard eigen decomposition for online GPR covariance update is cost-prohibitive, we employ incremental eigen decomposition in$\ell _{1}$-SRLSMF and SRLSMF GPR methods. Simulations were conducted to evaluate the performance of our proposed$\ell _{1}$-SRLSMF and SRLSMF GPR methods to compensate for the detrimental impacts of excessive delay and packet loss associated with 5G-enabled Tactile Internet remote robotic surgery. The results demonstrate that our proposed framework can outperform state-of-the-art approaches in terms of haptic data generalization performance. Finally, we assess the proposed framework’s ability to meet the Tactile Internet requirement for remote robotic surgery and discuss future research directions.

With the explosive demands for data, content delivery networks are facing ever-increasing challenges to meet end-users' quality-of-experience requirements, especially in terms of delay. Content can be migrated from surrogate servers to local caches closer to end-users to address delay challenges. Unfortunately, these local caches have limited capacities, and when they are fully occupied, it may sometimes be necessary to remove their lower-priority content to accommodate higher-priority content. At other times, it may be necessary to return previously removed content to local caches. Downloading this content from surrogate servers is costly from the perspective of network usage, and potentially detrimental to the end-user QoE in terms of delay. In this paper, we consider an edge content delivery network with vehicular nodes and propose a content migration strategy in which local caches offload their contents to neighboring edge caches whenever feasible, instead of removing their contents when they are fully occupied. This process ensures that more contents remain in the vicinity of end-users. However, selecting which contents to migrate and to which neighboring cache to migrate is a complicated problem. This paper proposes a deep reinforcement learning approach to minimize the cost. Our simulation scenarios realized up to a 70% reduction of content access delay cost compared to conventional strategies with and without content migration.

Multi-access edge computing (MEC) is an emerging paradigm to meet the rapidly growing computation demands of mobile applications. This paper investigates the performance gains of cooperative computation offloading for MEC enabled FiWi enhanced HetNets with capacity-limited backhaul links. After presenting the envisioned two-tier MEC architecture for a FiWi based networking infrastructure, we propose a simple but efficient offloading strategy, which relies on the flexible trilateral cooperation between end-device, edge servers, and the remote cloud. We then present an analytical framework to estimate the average response time and energy consumption of mobile users for various offloading scenarios with different wireless access modes (i.e., WiFi and 4G LTE-A). The presented analysis flexibly allows for incorporating both offloaded and conventional human-to-human (H2H) traffic of mobile users as well as fixed (wired) subscribers. Finally, we present our self-organization based mechanism, which enables mobile users to make suitable energy-delay trade-offs by jointly minimizing the average task execution time and energy consumption, using only their local information. The obtained results demonstrate the feasibility of the proposed cooperative self-organizing offloading strategy and its superior performance over schemes with MEC- or cloud-only offloading strategies.

With the advent of safe collaborative robots, their seamless integration into human teams as teammates is starting to gain steam as part of the vision of the emerging Tactile Internet. The Tactile Internet lies at the nexus of computerization, automation, and robotization. While necessary, low task execution time and ultra-reliable human-robot connectivity are not sufficient to unleash the full potential of the resultant human-agent-robot teamwork (HART) applications. In this paper, we propose a context- and self-aware HART-centric allocation scheme for both physical and digital tasks to coordinate the automation and augmentation of mutually beneficial human-machine coactivities while spreading ownership of robots across users over integrated fiber-wireless (FiWi) Tactile Internet infrastructures. In addition to realizing collective context-awareness via HART-centric task coordination, we aim at exploiting local self-awareness in order to improve the energy-delay performance of robots. Further, we present an analytical framework to estimate the packet transmission delay and human-robot connection reliability. Our results indicate that our proposed context- and self-aware HART-centric task coordination scheme obtains a low task execution time while minimizing the energy consumption and operational expenditures (OPEX) of mobile robots.

With the advent of semi-autonomous robotic assistance systems, their integration into human teams is starting to gain steam as part of the vision of human+machine hybrid activities. Unlike their fully autonomous counterparts, semi-autonomous robotic systems mainly rely on human assistance from time to time via teleoperation when human expertise is needed to accomplish a given task. As these robots will need to request human assistance via teleoperation, mapping these requests to human teleoperators stands as a difficult optimization problem. In this paper, after shedding some light on our envisioned FiWi enhanced network infrastructure and its role in realizing the emerging Tactile Internet, we formulate the problem of joint prioritized scheduling and assignment of delay-constrained teleoperation tasks to human operators with the objective to minimize the average weighted task completion time, maximum tardiness, and average operational expenditure (OPEX) per task. We then propose our context-aware prioritized scheduling and task assignment (CAPSTA) algorithm to achieve suitable trade-offs between the contradicting objectives of the problem. Further, to estimate the end-to-end packet delay of local and non-local teleoperation over FiWi enhanced networks, we develop our analytical framework, which flexibly allows for the coexistence of conventional human-to-human (H2H) and haptic human-to-machine (H2M) traffic.

A critical challenge for video-on-demand (VoD) services is to provide an entertainment service with minimum playback delay. A passive optical network (PON) that employs a high-speed optical fiber from an optical line terminal (OLT) to a number of optical network units (ONUs) can offer high bandwidth for multimedia applications (such as VoD services) in an access network. We propose two novel video streaming techniques, called OLT broadcasting with ONU fast patching (BFP) and prediction-based OLT broadcasting and ONU fast patching (PBFP). The BFP scheme utilizes the ONU fast patching scheme at each ONU, and proposes a heuristic algorithm to find near optimum solutions of related optimization problem so that the worst-case playback delay (WPD) is minimized. The PBFP adds a prediction level for video popularity to the BFP scheme and uses a seamless channel transition technique to seamlessly change the number of channels allocated to videos. We study the efficiency of the proposed schemes when they are used in a shared wavelength division multiplexed passive optical network (Shared-WDM-PON) that adapts the broadcast nature of a GPON's downstream wavelength to WDM-PONs. Our simulation results indicate that the proposed schemes can improve both WPD and average playback delay performance parameters.