Search

This paper introduces a dual-based algorithm framework for solving the regularized online resource allocation problems, which have potentially nonconcave cumulative rewards, hard resource constraints, and a nonseparable regularizer. Under a strategy of adaptively updating the resource constraints, the proposed framework only requests approximate solutions to the empirical dual problems up to a certain accuracy and yet delivers an optimal logarithmic regret under a locally second-order growth condition. Surprisingly, a delicate analysis of the dual objective function enables us to eliminate the notorious log-log factor in regret bound. The flexible framework renders renowned and computationally fast algorithms immediately applicable, for example, dual stochastic gradient descent. Additionally, an infrequent re-solving scheme is proposed, which significantly reduces computational demands without compromising the optimal regret performance. A worst-case square-root regret lower bound is established if the resource constraints are not adaptively updated during dual optimization, which underscores the critical role of adaptive dual variable update. Comprehensive numerical experiments demonstrate the merits of the proposed algorithm framework.

Model averaging is a widely adopted technique in federated learning (FL) that aggregates multiple client models to obtain a global model. Remarkably, model averaging in FL yields a superior global model, even when client models are trained with non-convex objective functions and on heterogeneous local datasets. However, the rationale behind its success remains poorly understood. To shed light on this issue, we first visualize the loss landscape of FL over client and global models to illustrate their geometric properties. The visualization shows that the client models encompass the global model within a common basin, and interestingly, the global model may deviate from the basin's center while still outperforming the client models. To gain further insights into model averaging in FL, we decompose the expected loss of the global model into five factors related to the client models. Specifically, our analysis reveals that the global model loss after early training mainly arises from i) the client model's loss on non-overlapping data between client datasets and the global dataset and ii) the maximum distance between the global and client models. Based on the findings from our loss landscape visualization and loss decomposition, we propose utilizing iterative moving averaging (IMA) on the global model at the late training phase to reduce its deviation from the expected minimum, while constraining client exploration to limit the maximum distance between the global and client models. Our experiments demonstrate that incorporating IMA into existing FL methods significantly improves their accuracy and training speed on various heterogeneous data setups of benchmark datasets. Code is available at https://github.com/TailinZhou/FedIMA. IEEE

Metaverse applications such as virtual reality (VR) content streaming, require optimal resource allocation strategies for mobile edge computing (MEC) to ensure a high-quality user experience. In contrast to online reinforcement learning (RL) algorithms, which can incur substantial communication overheads and longer delays, the majority of existing works employ offline-trained RL algorithms for resource allocation decisions in MEC systems. However, they neglect the impact of desynchronization between the physical and digital worlds on the effectiveness of the allocation strategy. In this paper, we tackle this desynchronization using a continual RL (CRL) framework that facilitates the resource allocation dynamically for MEC-enabled VR content streaming. We first design a digital twin-empowered edge computing (DTEC) system and formulate a quality of experience (QoE) maximization problem based on attention-based resolution perception. This problem optimizes the allocation of computing and bandwidth resources while adapting the attention-based resolution of the VR content. The CRL framework in DTEC enables adaptive online execution in a time-varying environment. We propose three variants of CRL, namely Continual Deep Deterministic Policy Gradient (CDDPG), Prioritized Experience Replay - CDDPG (PER-CDDPG), and Freshness Prioritized Experience Replay - CDDPG (FPER-CDDPG). We evaluate these algorithms, including two other benchmarks, using extensive experiments. FPER-CDDPG shows superior performance in terms of average latency, QoE, and successful delivery rate as well as meeting the hfQoE requirements over long-term execution while ensuring system scalability with the increasing number of users. IEEE

The Metaverse has emerged to extend our lifestyle beyond physical limitations. As essential components in the Metaverse, digital twins (DTs) are the real-time digital replicas of physical items. Multi-access edge computing (MEC) provides responsive services to the end users, ensuring an immersive and interactive Metaverse experience. While the digital representation (DT) of physical objects, end users, and edge computing systems is crucial in the Metaverse, the construction of these DTs and the interplay between them have not been well-investigated. In this article, we discuss the bidirectional reliance between the DT and the MEC system and investigate the creation of DTs of objects and users on the MEC servers and DT-assisted edge computing (DTEC). To ensure seamless handover among MEC servers and to avoid intermittent Metaverse services, we also explore the interaction between local DTECs on local MEC servers and the global DTEC on the cloud server due to the dynamic nature of network states (e.g., channel state and users' mobility). We investigate a continual learning framework for resource allocation strategy in local DTEC through a case study. Our strategy mitigates the desynchronization between physical-digital twins, ensures higher learning outcomes, and provides a satisfactory Metaverse experience. © 2018 IEEE.

With huge amounts of data generated from Internet of Things (IoT) devices, data-driven technologies are increasingly applied by firms to improve their IoT-based services in real time. To facilitate efficient utilization of the collected data, the design of data trading markets becomes crucial. Two important practical concerns are: 1) the data buyers arrive in a sequential and arbitrary manner and 2) a firm faces externalities when data is purchased by competing firms. In this article, we design a data trading marketplace for data buyers arriving dynamically in real time where the early arrived data buyers will exert negative externalities on the late arrivals within the same competition. Specifically, in market operations, when a data buyer arrives, it needs to submit a bid based on the price and the current externalities posted by the market operator. After receiving the bid, the market operator will announce the data allocation and payment as well as update the price. To construct the detailed market mechanism, we propose the allocation rule, payment rule, and price update method, which can be proven theoretically to guarantee the desirable properties, including incentive compatibility, individual rationality, revenue maximization, and computation efficiency. These theoretical conclusions are also validated via our numerical experiments. © 2014 IEEE.

Due to the massive computing demands of the Internet of Things, mobile edge computing (MEC) has been extensively investigated as a means of providing computation-intensive and latency-sensitive services at the network edge. With increasing density of base stations (BSs), users are simultaneously served by multiple BSs, leading to the multicell MEC environment. Intelligent reflecting surface (IRS) provides a promising solution for constructing the virtual Line-of-Sight (LoS) links between cell-edge users (CEUs) and BSs. In this article, we investigate the joint channel estimation and resource allocation in the IRS-aided multicell MEC system. Instead of assuming the perfect channel state information (CSI), we propose a three-phase channel estimation method to obtain the CSI. Our purpose is to minimize the total joint energy and latency cost (JELC) in terms of both task-execution latency and energy consumption in the IRS-aided multicell MEC problem by jointly optimizing the task offloading volume, precoding matrix, and IRS phase shifts. We propose a quadratically constrained program (QCP)-assisted proximal policy optimization (PPO) reinforcement learning algorithm with two modules (i.e., QCP optimizer and PPO agent) execute iteratively. The QCP optimizer is utilized to compute the offloading decision variables, and the PPO agent is adapted to determine the optimal channel precoding matrix and the phase shifts of IRS. Numerical results validate that our QCP-assisted PPO algorithm executes more rapidly than benchmarks. Moreover, the proposed QCP-assisted PPO algorithm delivers the best performance compared to benchmarks. Furthermore, the multicell IRS-aided MEC framework yields additional performance gains compared to those without IRS. © 2014 IEEE.

The explosive development of the Internet of Things (IoT) has led to increased interest in mobile edge computing (MEC), which provides computational resources at network edges to accommodate computation-intensive and latency-sensitive applications. Intelligent reflecting surfaces (IRSs) have gained attention as a solution to overcome blockage problems during the offloading uplink transmission in MEC systems. This paper explores IRS-aided multi-cell networks that enable servers to serve neighboring cells and cooperate to handle resource exhaustion. We aim to minimize the joint energy and latency cost by jointly optimizing the computation tasks, edge computing resources, user beamforming, and IRS phase shifts. The problem is decomposed into two subproblems - the MEC subproblem and the IRS communication subproblem - using the block coordinate descent (BCD) technique. The MEC subproblem is reformulated as a nonconvex quadratic constrained problem (QCP), while the IRS communication subproblem is transformed into a weight-sum-rate problem with auxiliary variables. We propose an efficient algorithm to alternately optimize the MEC resources and IRS communication variables. Numerical results show that our algorithm outperforms benchmarks and that multi-cell MEC systems achieve additional performance gains when supported by IRS. © 2017 IEEE.

Due to the influence of stochastic gradients, the existing algorithms suffer from slow convergence, noise explosion, and even failure to converge in practice, which motivates us to propose an accelerated algorithm to tackle these issues. Recognizing the potential of gradient, momentum, and conjugate gradient as promising search directions, we propose a 3-D acceleration algorithm, which uses a weighted combination of these three basis. Specifically, in order to analyze the dynamics of the discrete-time algorithm during the update process, we provide a general framework for approximating the discrete-time algorithm in the weak sense by a continuous-time stochastic differential equation. We exploit the continuous-time formulation together with Lyapunov drift optimization to derive novel adaptive step sizes, which effectively improve the performance of the algorithm in stabilizing noise and accelerating convergence. Extensive numerical experiments demonstrate the proposed algorithm’s superiority in convergence rate, computation complexity, and noise robustness compared to state-of-the-art baselines. IEEE

This letter explores the potential benefits of incorporating a bias voltage in the operation of an RF rectifier diode. Despite the challenges posed by network matching and additional power consumption, the study demonstrates efficiency enhancements across a range of input power levels. These enhancements offer significant advantages for active energy harvesters equipped with a controllable small bias source.

Federated learning allows multiple clients to collaboratively train a model without exchanging their data, thus preserving data privacy. Unfortunately, it suffers significant performance degradation due to heterogeneous data at clients. Common solutions involve designing an auxiliary loss to regularize weight divergence or feature inconsistency during local training. However, we discover that these approaches fall short of the expected performance because they ignore the existence of a vicious cycle between feature inconsistency and classifier divergence across clients. This vicious cycle causes client models to be updated in inconsistent feature spaces with more diverged classifiers. To break the vicious cycle , we propose a novel framework named Fed erated learning with F eature A nchors (FedFA). FedFA utilizes feature anchors to align features and calibrate classifiers across clients simultaneously. This enables client models to be updated in a shared feature space with consistent classifiers during local training. Theoretically, we analyze the non-convex convergence rate of FedFA. We also demonstrate that the integration of feature alignment and classifier calibration in FedFA brings a virtuous cycle between feature and classifier updates, which breaks the vicious cycle existing in current approaches. Extensive experiments show that FedFA significantly outperforms existing approaches on various classification datasets under label distribution skew and feature distribution skew.

A challenge of ambient backscatter communication (AmBC) systems is signal recovery because the transmitted information bits are embedded in the ambient RF signals and these are unknown and uncontrollable. To meet this challenge, averaging-based energy detectors are typically used but consequently the data rate is low and there is an error floor. Here we propose a new detection strategy based on the ratio between signals received from a multiple-antenna Reader. The advantage of using the ratio is that ambient RF signals are removed directly from the embedded signals without averaging and hence it can increase data rates and avoid the error floor. Different from the original ratio detector that uses the magnitude ratio of the signals between two Reader antennas, in our proposed approach, we utilize the complex ratio so that phase information is preserved and propose an accurate linear channel model approximation. This allows the application of existing linear detection techniques from which we can obtain a minimum distance detector and closed-form expressions for bit error rate (BER). Methods for the estimation of channel state information (CSI) are also provided. In addition, coding and interleaving are also included to further enhance the BER. The results are also general, allowing any number of Reader antennas to be utilized in the approach. Numerical results demonstrate the proposed approach performs better than approaches based on energy detection and the original ratio detectors. IEEE

Although deep neural networks (DNNs) have been remarkably successful in numerous areas, the performance of DNN is compromised in federated learning scenarios because of the large model size. A large model can induce huge communication overhead during the federated training, and also induce infeasible storage and computation burden at the clients during the inference. To address these issues, we investigate structured model compression in federated learning to construct sparse models with regular structure such that they require significantly less communication, storage and computation resources. We do this by proposing a three-layer hierarchical prior, which can promote a common regular sparse structure in the local models. We design a decentralized Turbo variational Bayesian inference (D-Turbo-VBI) algorithm to solve the resulting federated training problem. With the common regular sparse structure, both upstream and downstream communication overhead can be reduced, and the final model also has a regular sparse structure, which requires significantly less local storage and computation resources. Simulation results demonstrate that our proposed algorithm can efficiently reduce the communication overhead during federated training and the resulting model can achieve a significantly lower sparsity rate and inference time compared to the baselines while maintaining a competitive accuracy. IEEE

The continuously growing size of deep neural networks (DNNs) has sparked a surge in research on model compression techniques. Among these techniques, multi-resolution model compression has emerged as a promising approach which can generate multiple DNN models with shared weights and different computational complexity (resolution) through a single training. However, in most existing multi-resolution compression methods, the model structures for different resolutions are either predefined or uniformly controlled. This can lead to performance degradation as they fail to implement systematic compression to achieve the optimal model for each resolution. In this paper, we propose to perform multi-resolution compression from a Bayesian perspective. We design a resolution-aware likelihood and a two-layer prior for the channel masks, which allow joint optimization of the shared weights and the model structure of each resolution. To solve the resulted Bayesian inference problem, we develop a low complexity partial update block variational Bayesian inference (PUB-VBI) algorithm. Furthermore, we extend our proposed method into the arbitrary resolution case by proposing an auxiliary neural network (NN) to learn the mapping from the input resolution to the corresponding channel masks. Simulation results show that our proposed method can outperform the baselines on various NN models and datasets.

Mobile edge computing (MEC) resource allocation for remote rendering in virtual reality (VR) content streaming is critical for user experience. However, resource allocation becomes challenging due to the desynchronization between the physical and digital worlds in digital twin-empowered MEC. This paper presents our continual RL framework that facilitates dynamic resource allocation for MEC-enabled VR content streaming. We first design a digital twin-empowered edge computing (DTEC) system and formulate a maximization problem that considers attention-based resolution perception to maximize the quality of experience (QoE). This problem optimizes the allocation of computing and bandwidth resources while adapting the attention-based resolution of the VR content. We then apply continual reinforcement learning (CRL) to enable adaptive attention-based resolution VR streaming in a time-varying environment. We base the CRL's reward function on the QoE and horizon-fairness QoE (hfQoE) constraints. We support CRL with prioritized experience replay - continual deep deterministic policy gradient (PER-CDDPG) to enhance the performance of continual learning in the presence of time-varying DT updates. We test PER-CDDPG using extensive experiments and evaluation. PER-CDDPG outperforms the benchmarks in terms of average latency, QoE, and successful delivery rate as well as meeting the hfQoE requirements and performance over long-term execution while ensuring system scalability with the increasing number of users.

Automated algae classification using machine learning is a more efficient and effective solution compared to manual classification, which can be tedious and time-consuming. However, the practical application of such a classification approach is restricted by the scarcity of labeled freshwater algae datasets, especially for rarer algae. To overcome these challenges, this study proposes to generate artificial algal images with StyleGAN2-ADA and use both the generated and real images to train machine-learning-driven algae classification models. This approach significantly enhances the performance of classification models, particularly in their ability to identify rare algae. Overall, the proposed approach improves the F1-score of lightweight MobileNetV3 classification models covering all 20 freshwater algae covered in this research from 88.4% to 96.2%, while for the models that cover only the rarer algae, the experiments show an improvement from 80% to 96.5% in terms of F1-score. The results show that the approach enables the trained algae classification systems to effectively cover algae with limited image data.

Identifying and quantifying algal genera in images are crucial for understanding their ecological impact. Algal data are often imbalanced, limiting detection model accuracy. This paper presents a novel data augmentation method using StyleGAN2-ADA to enhance algal image instance segmentation. StyleGAN2-ADA generates artificial single-algal images to address data scarcity and imbalance. We train a Cascaded Mask R-CNN with Swin Transformer on a combined data set of real and artificial multigenera algal images and evaluate performance using the COCO mAP metric. The approach improves bounding box detection performance by 17.9% on all genera and 32.1% on rare genera compared with the baseline model. Additionally, 50% more artificial data yield significant enhancements without excessive artificial data use. The GAN-based augmentation technique shows a performance improvement in both Swin-Tiny and ResNet-50 backbone models, suggesting adaptability for various machine learning models. The increased mAP leads to the accurate identification of harmful algae genera, allowing for better prevention and mitigation. This method offers a superior data augmentation solution for accurate algal instance segmentation and can benefit applications challenged by imbalanced and scarce data.

In this paper, we propose a new solution to improve the achievable rate of radio frequency (RF) powered cognitive radio networks (CRNs) with ambient backscatter communication (AmBC). Assisted with AmBC, the secondary transmitter (ST) can harvest energy and backscatter ambient signals when the primary channel is busy, which enhances the achievable rate compared with conventional RF-powered CRNs adopting the harvest-then-transmit (HTT) protocol. Our work proposes an RF-powered CRN that uses a multi-antenna ST since implementing multiple antennas on ST can enhance energy harvesting and increase the data rate. We discuss the corresponding time sharing and antenna selection tradeoffs and propose a low-complexity and time-efficient block coordinate descent (BCD)-assisted exhaustive search algorithm to find the optimal tradeoff that maximizes the data rate of the system. Simulation results show that our proposed scheme outperforms both the HTT mode and the ambient backscatter technique, leading to improved overall system performance. © 2023 IEEE.

Multiple-input multiple-output (MIMO) ambient backscatter communication (AmBC) systems are investigated in order to develop approaches to achieve power, multiplexing, and diversity gains. These results can be utilized to motivate the development of MIMO AmBC by providing performance bounds on the MIMO AmBC gains. Our approach to the investigation is to reformulate the MIMO AmBC channel model as an accurate linear MIMO channel model. Using this model, we show that the MIMO AmBC received signal is real-valued, hence the dimension of the received signal is halved. In addition, we show that MIMO AmBC has a per antenna constant modulus transmit signal constraint. Therefore, increases in antennas at the Tag provide a power gain in contrast to conventional MIMO systems. Under these constraints, when channel state information at the Reader (CSIR) is available, we provide estimates of channel capacity. Assuming channel state information at the Tag (CSIT) is also available, we use a fixed-point iteration algorithm to maximize channel capacity. With CSIT, beamforming design and the corresponding majorization-minimization (MM) algorithm are proposed to find the optimal transmit and receive beamformers so that diversity gain can be leveraged. It is also shown that in the low signal-to-noise ratio (SNR) operating region, beamforming design maximizes channel capacity. In addition, we provide numerical results for the diversity-multiplexing tradeoff (DMT). Utilizing comparisons between AmBC and conventional MIMO, we highlight the unique characteristics of MIMO AmBC. These approaches to maximize power, capacity, and diversity gains demonstrate the potential of MIMO AmBC.

The cloud-based solutions are becoming inefficient due to considerably large time delays, high power consumption, and security and privacy concerns caused by billions of connected wireless devices and typically zillions of bytes of data they produce at the network edge. A blend of edge computing and Artificial Intelligence (AI) techniques could optimally shift the resourceful computation servers closer to the network edge, which provides the support for advanced AI applications (e.g., video/audio surveillance and personal recommendation system) by enabling intelligent decision making on computing at the point of data generation as and when it is needed, and distributed Machine Learning (ML) with its potential to avoid the transmission of the large dataset and possible compromise of privacy that may exist in cloud-based centralized learning. Besides, the deployment of AI techniques to redesign end-to-end communication is attracting attention to improve communication performance. Therefore, the interaction of AI and wireless communications generates a new concept, named native AI wireless networks. In this paper, we conduct a comprehensive overview of recent advances in distributed intelligence in wireless networks under the umbrella of native AI wireless networks, with a focus on the design of distributed learning architectures for heterogeneous networks, on AI-enabled edge computing, on the communication-efficient technologies to support distributed learning, and on the AI-empowered end-to-end communications. We highlight the advantages of hybrid distributed learning architectures compared to state-of-the-art distributed learning techniques. We summarize the challenges of existing research contributions in distributed intelligence in wireless networks and identify potential future opportunities. © 2020 IEEE.

The online knapsack problem is a classic online resource allocation problem in networking and operations research. Its basic version studies how to pack online arriving items of different sizes and values into a capacity-limited knapsack. In this paper, we study a general version that includes item departures, while also considering multiple knapsacks and multi-dimensional item sizes. We design a threshold-based online algorithm and prove that the algorithm can achieve order-optimal competitive ratios. Beyond worst-case optimized algorithms, we also propose a data-driven online algorithm that can achieve near-optimal average performance under typical instances while guaranteeing the worst-case performance.

With the growth of neural network size, model compression has attracted increasing interest in recent research. As one of the most common techniques, pruning has been studied for a long time. By exploiting the structured sparsity of the neural network, existing methods can prune neurons instead of individual weights. However, in most existing pruning methods, surviving neurons are randomly connected in the neural network without any structure, and the non-zero weights within each neuron are also randomly distributed. Such irregular sparse structure can cause very high control overhead and irregular memory access for the hardware and even increase the neural network computational complexity. In this paper, we propose a three-layer hierarchical prior to promote a more regular sparse structure during pruning. The proposed three-layer hierarchical prior can achieve per-neuron weight-level structured sparsity and neuron-level structured sparsity. We derive an efficient Turbo-variational Bayesian inferencing (Turbo-VBI) algorithm to solve the resulting model compression problem with the proposed prior. The proposed Turbo-VBI algorithm has low complexity and can support more general priors than existing model compression algorithms. Simulation results show that our proposed algorithm can promote a more regular structure in the pruned neural networks while achieving even better performance in terms of compression rate and inferencing accuracy compared with the baselines.

By employing powerful edge servers for data processing, mobile edge computing (MEC) has been recognized as a promising technology to support emerging computation-intensive applications. Besides, non-orthogonal multiple access (NOMA)-aided MEC system can further enhance the spectral efficiency with massive tasks offloading. However, with more dynamic devices brought online and the uncontrollable stochastic channel environment, it is even desirable to deploy appealing technique, i.e., intelligent reflecting surfaces (IRS), in the MEC system to flexibly tune the communication environment and improve the system energy efficiency. In this paper, we investigate the joint offloading, communication and computation resource allocation for the IRS-assisted NOMA MEC system. We first formulate a mixed integer energy efficiency maximization problem with system queue stability constraint. We then propose the Lyapunov-function-based Mixed Integer Deep Deterministic Policy Gradient (LMIDDPG) algorithm which is based on the centralized reinforcement learning (RL) framework. To be specific, we design the mixed integer action space mapping which contains both continuous mapping and integer mapping. Moreover, the award function is defined as the upper-bound of the Lyapunov drift-plus-penalty function. To enable end devices (EDs) to choose actions independently at the execution stage, we further propose the Heterogeneous Multi-agent LMIDDPG (HMA-LMIDDPG) algorithm based on distributed RL framework with homogeneous EDs and heterogeneous base station (BS) as heterogeneous multi-agent. Numerical results show that our proposed algorithms can achieve superior energy efficiency performance to the benchmark algorithms while maintaining the queue stability. Specially, the distributed structure HMA-LMIDDPG can acquire more energy efficiency gain than the centralized structure LMIDDPG.

An electric vehicle based mobility-on-demand (EMoD) system provides shared transportation (e.g., car-sharing or ride-sharing) to satisfy customers' individual mobility demands. It has been recognized as a vital alternative form of transportation between public and private transportations in future sustainable cities. Constrained by the long charging time and limited driving range of EVs, an operator of an EMoD system demands for decision-making models and algorithms to manage the mobility and energy of EVs to best serve customers with least costs. In this paper, we propose a stochastic dynamic program (DP) to model three operational decisions of the EMoD system: i) dispatching EVs to serve mobility demand from customers, ii) repositioning EVs to accommodate the unbalanced mobility demands between service regions, and iii) recharging EVs to maintain their sufficient state-of-charge levels. To handle this large-scale DP problem, we first observe and prove that it has a coordinate-wise concave value function. Based on this structural property, we propose to use a separable piecewise linear function to approximate the value function and design an approximation-based algorithm to efficiently derive the decision policy. Numerical tests show that our proposed algorithm significantly outperforms the existing model-free approaches (e.g., greedy heuristic and Q-learning) that fail to take into account the structural properties of the DP problem.

The stochastic economic dispatch (SED) provides a promising solution for short-term cost-efficiency with high renewable penetration. However, the long-term efficiency of SED, i.e., promoting efficient investments in both flexibility resources and renewable generations, has rarely been studied. We theoretically reveal the financial incentives for efficient investments in SED from both perspectives of the short-term and long-term performances: 1) We demonstrate the rationality behind the short-term failure of cost recovery or revenue adequacy, which provides proper signals to corresponding dispatchable generators, renewable generators, or system operator for long-term reconstruction and installation. 2) Through theoretical analysis of the statistical properties, we prove that the long-term expected profits of renewable and dispatchable generators provide appropriate signals to guide their future investments. In conclusion, the SED can indeed motivate the investments of flexibility resources and renewables to match up better with each other in the network, benefiting the system efficiency and reliability in the long run. Simulations are conducted for validation. IEEE

In this article, we consider an online version of the well-studied network utility maximization problem, where users arrive one by one and a network operator makes irrevocable rate allocation decisions for each user without knowing the details of future arrivals. We propose a threshold-based algorithm and analyze its worst-case performance. We prove that the competitive ratio of the proposed algorithm is logarithmic in the maximum number of links requested by a user. Extensive simulations are conducted to demonstrate the performance advantage of our proposed algorithm in comparison with two state-of-the-art algorithms. In addition, we devise an adaptive implementation of our algorithm with online learning.

Federated learning (FL) allows multiple clients to train a global model while keeping data locally. It has been well recognized that FL suffers from data heterogeneity, leading to drifts in client updates and deteriorating global model performance. However, there is a lack of research on the relationship between client and global models in the parametric space, leaving unclear where client updates end their drifts. To bridge this gap, we consider different model solutions to the client and FL objectives (referred to as client modes and global modes, respectively) and use the mode connectivity of neural networks to measure model performance change along different parametric paths. Specifically, we conduct comprehensive studies on the linear and non-linear connectivity of client modes and global modes to reveal their geometric relationship in the solution space. We find that as data heterogeneity reduces in FL, the connectivity (e.g., performance change) between client and global modes on different paths becomes more similar. This results in more overlapping low-error solutions between client and FL objectives in the solution space. Furthermore, when two global modes are linearly connected, a connectivity barrier emerges, which weakens as heterogeneity decreases. This barrier disappears and becomes independent of data heterogeneity when nonlinear mode connectivity is considered. Our empirical results demonstrate that client and global modes are close to each other in the solution space while they differ in connectivity under varying data heterogeneity.

Non-orthogonal multiple access (NOMA)-aided mobile edge computing (MEC) system can enhance the spectral-efficiency with massive tasks offloading. However, with more dynamic devices and the uncontrollable stochastic channel environment, it is even desirable to deploy appealing technique, i.e., intelligent reflecting surfaces (IRS), in the MEC system to flexibly adjust the communication environment and improve the system energy-efficiency. In this paper, we investigate the joint offloading, communication and computation resource allocation for IRS-assisted NOMA-aided MEC system. We firstly formulate a mixed integer energy-efficiency maximization problem with the system queue stability constraint. We then propose a Het-erogeneous Multi-agent Lyapunov-function-based Mixed Integer Deep Deterministic Policy Gradient (HMA-LMIDDPG) algorithm which is based on the multi-agent reinforcement learning (MARL) framework with homogeneous edge devices (EDs) and heterogeneous base station (BS) as heterogeneous multi-agent. Numerical results show that our proposed algorithms can achieve superior energy-efficiency performance to the benchmark algorithms while maintaining the queue stability. © 2023 IEEE.

The online knapsack problem is a classic online resource allocation problem in networking and operations research. Its basic version studies how to pack online arriving items of different sizes and values into a capacity-limited knapsack. In this paper, we study a general version that includes item departures, while also considering multiple knapsacks and multi-dimensional item sizes. We design a threshold-based online algorithm and prove that the algorithm can achieve order-optimal competitive ratios. Beyond worst-case performance guarantees, we also aim to achieve near-optimal average performance under typical instances. Towards this goal, we propose a data-driven online algorithm that learns within a policy-class that guarantees a worst-case performance bound. In trace-driven experiments, we show that our data-driven algorithm outperforms other benchmark algorithms in an application of online knapsack to job scheduling for cloud computing. © 2022 ACM.

Gas-fired generation has been considered as one of the mainstream generation technologies nowadays and becomes a vital part for the decarbonization journey in Hong Kong. Having a significant impact to the business and environment, the utilization of generation units has to be optimized in such a way that additional gas generation can be accommodated and the performance of efficiency and reliability for the existing eight gas-fired generation units can be enhanced. Currently, at Black Point Power Station (BPPS) in Hong Kong, the generation scheduling is generally done manually based on traditional approach. Each unit has its specific characteristic in terms of efficiency, cost, flexibility, reliability and performance according to their individual hardware condition, degradation and running regime. As those cost and various machine factors are not considered effectively, efficiently and intelligentially, the overall generation costs as well as the carbon emission reduction have not optimized as a result. This paper is to develop a unit commitment optimization (UCO) model with help of the state of art artificial intelligence and machine learning for optimizing the daily scheduling of eight gas-fired generation units at BPPS. It illustrates the comparison and advantage of the newly developed optimization model with the existing method. Moreover, a holistic review on the real-life generation cost, efficiency, maximum output power and condition related data is completed to minimize the uncertainty of modelling. All the factors that affecting optimal scheduling are further validated during the testing and tuning phase. The simulation result shows that the model is practical with sound accuracy in efficiency and cost optimization for the BPPS machine scheduling.

Resource-constrained end devices can offload computation to backend clouds. The stochastic wireless channel that an end device is connected to can introduce bursty computation demand to the cloud. Specifically, under good channel conditions, a device can transmit more data to the cloud, which consequently yields higher instantaneous computation demand. Conversely, poor channel conditions can result in lower instantaneous demand. The performance indicator for such a mobile cloud computing system is the average of the response time, which is the time span from the arrival of the computation demand at the backend cloud instance to the completion of its execution. The question we target in this paper is how resources should be provisioned for the backend cloud instance to address this bursty computation demand and guarantee a desired quality-of-service (QoS), namely, a user-specified average response time. To answer this question, we model the mobile cloud computing system as two tandem queues. We analyze this queueing network using the fluid flow analysis framework, and derive the analytical relationship between the required resource capacity at the backend cloud instance and the desired QoS, given the workload generation process at the end device and the wireless channel conditions. Having obtained the required resource capacity for a desired QoS, we then determine whether it is economical to provision this resource capacity by subscribing to the traditional static instance or the recently introduced burstable instance offered by public cloud providers. Finally, trace-driven simulations validate our theoretical results.

In this paper, we consider a parking lot that manages the charging processes of its parked electric vehicles (EVs). Upon arrival, each EV requests a certain amount of energy. This request should be fulfilled before the EV’s departure. It is of critical importance to coordinate the EVs’ charging rates to smooth out the load profile of the parking lot because inappropriate charging rates can lead to sharp spikes and fluctuations on the load profile, imposing negative effects on the power grid. Meanwhile, empirical studies show that many parking lots exhibit statistical patterns on EV dynamics. For example, the bulk of EVs arrive during rush hours. Therefore, in this paper, we incorporate such patterns into charging rate coordination. Although the statistical patterns can be summarized from historical data, they are difficult to be analytically modeled. As a result, we adopt a model-free deep reinforcement learning approach. We also take the latest continuous charging rate control technology into consideration. The decision variables are thus continuous and a policy gradient algorithm is needed to perform reinforcement learning. Technically, we first formulate the problem as a Markov decision process (MDP) with unknown state transition probabilities. To further derive a deep policy gradient algorithm, the challenge lies in the inconsistent and state-dependent action space of the MDP model, due to the constraint to satisfy EVs’ energy demands before their scheduled departure. To tackle the challenge, we design a customized model for neural network training by extending the action space to be consistent and state-independent, and revise the reward function to penalize the neural network output if it is beyond the action space of the original MDP model. With this customized model, we then develop a deep policy gradient algorithm based on the proximal policy gradient framework. Numerical results show that our algorithm outperforms the benchmarks. IEEE

With the popularization of ride-sharing transportation and increasing penetration of electric vehicles (EVs) in recent years, the electric mobility-on-demand (EMoD) system is emerging as a promising means to provide ride-sharing services in the context of sustainable cities. In this paper, we focus on sequential decision-making for the operator of an EMoD system by considering both the passengers' utility and system revenue. Specifically, we design a pricing mechanism to incentivize passengers with spatially and temporally different demand to make different mobility choices. After the passengers' demand is realized, the operator makes operational decisions, including dispatching and repositioning EVs between service regions, and recharging EVs to maintain their energy levels. Therefore, a bi-level and dynamic programming problem is formulated to model these decisions. To solve this problem, we first transform the bi-level problem into a single-level one based on the structural properties of the formulation. Furthermore, we rigorously prove the coordinate-wise concavity of the single-level formulation and efficiently obtain near-optimal solutions based on approximation. Numerical tests show that the proposed dynamic pricing mechanism achieves a significantly better performance than static pricing and other existing model-free approaches (e.g., Q-learning).

In this article, we study an energy-efficient nonorthogonal multiple access (NOMA) enabled multiaccess edge computing (MEC) system with strict latency requirements. We aim to minimize the energy consumption of all users by optimizing the resource allocation (including power and computation resources) and subchannel assignment, subject to the given latency constraint. The formulated problem, however, is a nonconvex combinatorial optimization problem. Nevertheless, we decompose the problem into a resource allocation subproblem and a subchannel assignment subproblem, and then solve the two subproblems iteratively. On one hand, we investigate the hidden convexity of the resource allocation subproblem under the optimal conditions, and propose an efficient algorithm to optimally allocate the resources by dual decomposition methods. On the other hand, we formulate the subchannel assignment subproblem into an integer linear programming problem and strictly prove that the problem is nondeterministic polynomial-time hard. We then solve it optimally by branch-and-bound methods, which is shown to be efficient in extensive simulations. Moreover, through considerable simulation results, we show that our proposed algorithm helps greatly reduce users’ energy consumption when communication resources (e.g., bandwidth) are limited. Additionally, it is verified that NOMA outperforms orthogonal multiple access in multiuser latency-sensitive MEC systems.

Gas-fired generation has been considered one of the mainstream generation technologies nowadays and has become a vital part of the decarbonization journey in Hong Kong. Having a significant impact on the business and environment, the utilization of generation units has to be optimized so that additional gas generation can be accommodated and the performance of efficiency and reliability for the existing eight gas-fired generation units can be enhanced. Currently, at Black Point Power Station (BPPS) in Hong Kong, the generation scheduling is done manually based on the traditional approach. Each unit has its specific characteristic in terms of efficiency, cost, flexibility, reliability and performance according to its individual hardware condition, degradation and running regime. As those costs and various machine factors are not considered effectively, efficiently and intelligentially, the overall generation costs, as well as the carbon emission reduction, have not been optimized as a result. This thesis is to develop a unit commitment optimization (UCO) model with the help of state-of-the-art artificial intelligence and machine learning for optimizing the daily scheduling of eight gas-fired generation units at BPPS. It illustrates the comparison and advantage of the newly developed optimization model with the existing method. Moreover, a holistic review of the real-life generation cost, efficiency, maximum output power and condition-related data is completed to minimize modelling uncertainty. All the factors that affect the scheduling are further validated during the testing and tuning phase. The simulation result shows that the model is practical with sound good accuracy in efficiency, and cost optimization for the BPPS machine scheduling.