Quality of Service (QoS) routing is a critical technique for delivering differentiated services under limited network resources to meet the specific requirements of endpoint applications. Despite the development of various routing algorithms and management architectures, achieving fine-grained QoS optimization within existing networks remains challenging due to the lack of endpoint control over routing decisions. This paper introduces PanQoSR, the first application of path-aware networks (PAN) in QoS routing. PanQoSR leverages the inherent capabilities of PAN by offloading path computation and selection to the endpoint, enabling flow-level fine-grained QoS optimization. PanQoSR addresses several key challenges in applying PAN to QoS routing. First, by leveraging existing network technologies and protocols, PanQoSR remains fully compatible with legacy networks without requiring significant modifications. Second, by introducing an ε−Constraint Pathfinding (ε−CP) algorithm for intra-AS path computation and a Nonlinear Cost Pathfinding (NCP) algorithm for inter-AS path computation, PanQoSR achieves both high QoS guarantees and computational efficiency. Experimental results show that PanQoSR reduces QoS violation rates by up to 70.8% compared to baselines, while also decreasing inter-AS path computation time by 22.4% to 65.6%.

High-definition (HD) cameras for surveillance and road traffic have experienced tremendous growth, demanding intensive computation resources for real-time analytics. Recently, offloading frames from the front-end device to the back-end edge server has shown great promise. In multi-stream competitive environments, efficient bandwidth management and proper scheduling are crucial to ensure both high inference accuracy and high throughput. To achieve this goal, we propose BiSwift, a bi-level framework that scales the concurrent real-time video analytics by a novel adaptive hybrid codec integrated with multi-level pipelines, and a global bandwidth controller for multiple video streams. The lower-level front-back-end collaborative mechanism (called adaptive hybrid codec) locally optimizes the accuracy and accelerates end-to-end video analytics for a single stream. The upper-level scheduler aims to accuracy fairness among multiple streams via the global bandwidth controller. The evaluation of BiSwift shows that BiSwift is able to real-time object detection on 9 streams with an edge device only equipped with an NVIDIA RTX3070 (8G) GPU. BiSwift improves 10%~21% accuracy and presents$1.2\sim 9\times $throughput compared with the state-of-the-art video analytics pipelines.

Forecasting stock prices is complex and challenging. Uncovering correlations among stocks has proven to enhance stock price forecasting. However, existing correlation discovery methods, such as concept-based methods, are slow, inaccurate, and limited by their reliance on predefined concepts and manual analysis. In this paper, we propose COGRASP, a novel approach for stock price forecasting that constructs stock co-occurrence graphs automatically by analyzing rapidly updated sources such as reports, newspapers, and social media. Besides, we aggregate forecasts across multiple timescales (i.e., long-, medium-, and short-term) to capture multi-timescale trends fluctuations, thereby enhancing price forecasting accuracy. In experiments with real-world open-source stock market data, COGRASP outperforms state-of-the-art methods.

Waypoint planning is critical for Unmanned Aerial Vehicle (UAV) operations, particularly for surveillance and monitoring applications. With the rapid development and deployment of UAVs, an increasing number of industrial UAVs are equipped with multiple cameras to enhance monitoring capabilities and operational efficiency. Meanwhile, a new type of camera supporting adjustable pan and zoom is emerging and rapidly being deployed. While UAVs equipped with multi-adjustable cameras enhance flexibility and precision in capturing dynamic scenes, they also introduce new challenges in optimizing waypoints to ensure efficient coverage and accurate data collection. In this paper, we propose WiPlan, which aims to determine the optimal UAV waypoints while dynamically adjusting the pan (horizontal rotation) and zoom (focal length) of cameras to maximize overall monitoring utility. This problem involves two coupled NP-hard problems, making it significantly more complex to solve compared to previous work. In tackling this challenge, we construct WiPlan as a two-level optimization problem. The results show that our algorithm improves monitoring utility by at least$1.73 \times$compared to state-of-the-art algorithms. Moreover, we test WiPlan using a real-world outdoor field, which includes 23 objects and a two-camera UAV with 7 waypoints. The results demonstrate that our algorithm successfully monitors 65 % of the maximum possible monitored targets, outperforming the baselines by factors of$4.25 \times$and$15 \times$, respectively.

Access networks, even with advanced 5G technology, often face bottlenecks when supporting concurrent real-time Ultra High Definition (UHD) video streams with high bandwidth and low latency (e.g., under 10 ms of one-way delay) requirements. Traditionally, end systems employ a combination of flow and congestion control mechanisms to control the sending rate to avoid overwhelming the receiver and the network. However, such control efforts induce prolonged tail delays, thereby sharply reducing the number of UHD video streams meeting delivery deadlines, and sometimes even zero. These outcomes are largely due to the inaccurate network status estimation associated with the control mechanisms. To address this challenge, we propose CFC, a deadline-oriented flow control mechanism that employs cross-layer status estimation to maximize user satisfaction with deadlines. CFC accurately assesses cross-layer information, including flow status and 5G access network status at minimal expense, thus ensuring the deadlines through effective concurrent flow control. Our experiments, conducted in both simulation and testbed settings, demonstrate significant improvements in delay and load-balancing for both reliable and unreliable transmissions.

High-definition (HD) cameras for surveillance and road traffic have experienced tremendous growth, demanding intensive computation resources for real-time analytics. Recently, offloading frames from the front-end device to the back-end edge server has shown great promise. In multi-stream competitive environments, efficient bandwidth management and proper scheduling are crucial to ensure both high inference accuracy and high throughput. To achieve this goal, we propose BiSwift, a bi-level framework that scales the concurrent real-time video analytics by a novel adaptive hybrid codec integrated with multi-level pipelines, and a global bandwidth controller for multiple video streams. The lower-level front-back-end collaborative mechanism (called adaptive hybrid codec) locally optimizes the accuracy and accelerates end-to-end video analytics for a single stream. The upper-level scheduler aims to accuracy fairness among multiple streams via the global bandwidth controller. The evaluation of BiSwift shows that BiSwift is able to real-time object detection on 9 streams with an edge device only equipped with an NVIDIA RTX3070 (8G) GPU. BiSwift improves 10%∼21% accuracy and presents 1.2∼ 9× throughput compared with the state-of-the-art video analytics pipelines.

Abstract Video streaming from cameras to backend cloud or edge servers for neural-based analytics has gained significant popularity. However, the transmission of data from cameras to a backend raises substantial privacy concerns, particularly regarding sensitive information like facial data. To offer privacy protection, visual processing techniques, such as Generative Adversarial Networks (GANs), have been employed on cameras to blur and safeguard such data intelligently. However, these techniques frequently face memory challenges, particularly when dealing with high-resolution videos. In this paper, we propose PIMO, a memory-efficient visual privacy protection scheme designed to effectively blur video content leveraging adaptive slicing of frames and resolution degradation. Our extensive experimental evaluations validate that PIMO’s adaptive mechanism proficiently navigates fluctuating memory constraints. Furthermore, utilizing a content-based blur scheme, our approach can maintain an impressive mean precision of 95.2%, as compared to the original, non-blurred images.

Cross-modality data translation has attracted great interest in medical image computing. Deep generative models show performance improvement in addressing related challenges. Nevertheless, as a fundamental challenge in image translation, the problem of zero-shot learning cross-modality image translation with fidelity remains unanswered. To bridge this gap, we propose a novel unsupervised zero-shot learning method called Mutual Information guided Diffusion Model, which learns to translate an unseen source image to the target modality by leveraging the inherent statistical consistency of Mutual Information between different modalities. To overcome the prohibitive high dimensional Mutual Information calculation, we propose a differentiable local-wise mutual information layer for conditioning the iterative denoising process. The Local-wise-Mutual-Information-Layer captures identical cross-modality features in the statistical domain, offering diffusion guidance without relying on direct mappings between the source and target domains. This advantage allows our method to adapt to changing source domains without the need for retraining, making it highly practical when sufficient labeled source domain data is not available. We demonstrate the superior performance of MIDiffusion in zero-shot cross-modality translation tasks through empirical comparisons with other generative models, including adversarial-based and diffusion-based models. Finally, we showcase the real-world application of MIDiffusion in 3D zero-shot learning-based cross-modality image segmentation tasks.

The quality of the video stream is the key to neural network-based video analytics. However, low-quality video is inevitably collected by existing surveillance systems because of poor-quality cameras or over-compressed/pruned video streaming protocols, e.g., as a result of upstream bandwidth limit. To address this issue, existing studies use quality enhancers (e.g., neural super-resolution) to improve the quality of videos (e.g., resolution) and eventually ensure inference accuracy. Nevertheless, directly applying quality enhancers does not work in practice because it will introduce unacceptable latency. In this paper, we present AccDecoder, a novel accelerated decoder for real-time and neural-enhanced video analytics, selects a few frames adaptively via Deep Reinforcement Learning (DRL) to enhance the quality and inference then reuse on the unselected ones. Next, we extend AccDecoder to AccDecoder$+$by formulating the resolution-involved Markov decision process (MDP) to achieve resolution adaptation; it aims to trade accuracy and latency corresponding under various video resolutions. Proved by experiments, AccDecoder provides efficient inference capability via filtering important frames using DRL for DNN-based inference and reusing the results for the other frames via extracting the reference relationship among frames and blocks, which contributes 6-21% accuracy improvement and a latency reduction of 20-80% than baselines. Compared with AccDecoder, AccDecoder$+$achieves an additional 2-7% accuracy improvement.

Communication is supposed to improve multi-agent collaboration and overall performance in cooperative Multi-agent reinforcement learning (MARL). However, such improvements are prevalently limited in practice since most existing communication schemes ignore communication overheads (e.g., communication delays). In this paper, we demonstrate that ignoring communication delays has detrimental effects on collaborations, especially in delay-sensitive tasks such as autonomous driving. To mitigate this impact, we design a delay-aware multi-agent communication model (DACOM) to adapt communication to delays. Specifically, DACOM introduces a component, TimeNet, that is responsible for adjusting the waiting time of an agent to receive messages from other agents such that the uncertainty associated with delay can be addressed. Our experiments reveal that DACOM has a non-negligible performance improvement over other mechanisms by making a better trade-off between the benefits of communication and the costs of waiting for messages.

The quality of the video stream is key to neural network-based video analytics. However, low-quality video is inevitably collected by existing surveillance systems because of poor quality cameras or over-compressed/pruned video streaming protocols, e.g., as a result of upstream bandwidth limit. To address this issue, existing studies use quality enhancers (e.g., neural super-resolution) to improve the quality of videos (e.g., resolution) and eventually ensure inference accuracy. Nevertheless, directly applying quality enhancers does not work in practice because it will introduce unacceptable latency. In this paper, we present AccDecoder, a novel accelerated decoder for real-time and neural-enhanced video analytics. AccDecoder can select a few frames adaptively via Deep Reinforcement Learning (DRL) to enhance the quality by neural super-resolution and then up-scale the unselected frames that reference them, which leads to 6-21% accuracy improvement. AccDecoder provides efficient inference capability via filtering important frames using DRL for DNN-based inference and reusing the results for the other frames via extracting the reference relationship among frames and blocks, which results in a latency reduction of 20-80% than baselines.

Abstract Federated Learning (FL) suffers from the Non-IID problem in practice, which poses a challenge for efficient and accurate model training. To address this challenge, prior research has introduced clustered FL (CFL), which involves clustering clients and training them separately. Despite its potential benefits, CFL can be computationally and communicationally expensive when the data distribution is unknown beforehand. This is because CFL involves the entire neural networks of involved clients in computing the clusters during training, which can become increasingly time-consuming with large-sized models. To tackle this issue, this paper proposes an efficient CFL approach called LayerCFL that employs a Layer-wised clustering technique. In LayerCFL, clients are clustered based on a limited number of layers of neural networks that are pre-selected using statistical and experimental methods. Our experimental results demonstrate the effectiveness of LayerCFL in mitigating the impact of Non-IID data, improving the accuracy of clustering, and enhancing computational efficiency.

Obtaining accurate and timely estimates of socioeconomic status at fine geographical resolutions is essential for global sustainable development and the fight against poverty. However, data related to local socio-economic dynamics in rural villages is often either unavailable or outdated. To fill this gap, predicting local economic well-being with satellite imagery and machine learning has shown promising results. While state-of-the-art analyses currently mostly focus on predicting the levels of socio-economic status, finding temporal changes in rural villages’ economic well-being is essential for tracking the impacts of public policies (targeting e.g., poverty alleviation or access to various public services). In this paper, we propose an approach that utilizes pixel-wise differences in satellite images to classify temporal changes in average and median consumption expenditures (and income) in rural villages in Thailand and Vietnam between 2007 and 2017. We can distinguish between “Decline”, “Stagnation” and “Growth” in these outcomes with an F1 score of 58.8% using a Logistic Regression model. Regression-based approaches achieve an R2of up to 32.5% when predicting actual changes in these outcomes. Our approach demonstrates the feasibility of satellite-based estimates for measuring changes in local socio-economic dynamics.

In this article, we introduce a novel dynamic controller assignment algorithm targeting connected vehicle services and applications, also known as Internet of Vehicles (IoV). The proposed approach considers a hierarchically distributed control plane, decoupled from the data plane, and uses vehicle location and control traffic load to perform controller assignment dynamically. We model the dynamic controller assignment problem as a multi-agent Markov game and solve it with cooperative multi-agent deep reinforcement learning. Simulation results using real-world vehicle mobility traces show that the proposed approach outperforms existing ones by reducing control delay as well as packet loss.

Terrestrial infrastructure-based wireless networks do not always guarantee their resources will be shared uniformly by nodes in vehicular networks. This is due mainly to the uneven and dynamic geographical distribution of vehicles and path loss effects. In this paper, we leverage multiple fifth-generation (5G) unmanned aerial vehicles (UAVs) to enhance fairness in network resource allocation among vehicles by positioning UAVs on-demand as “flying communication infrastructure”. We propose a deep reinforcement learning (DRL) approach to determine UAVs’ position to improve network resource allocation fairness and efficiency while considering the UAVs’ flying range, communication range, and energy constraints. We use a parametric fairness function to attain a number of resource allocation objectives ranging from maximizing the total throughput of vehicles, maximizing minimum throughput, and achieving proportional bandwidth allocation. Simulation results show that the proposed DRL approach to UAV positioning can improve network resource allocation according to the targeted fairness objective.

With the exploitation of wind power, more turbines will be deployed at remote areas possibly with harsh working conditions (e.g., offshore wind farm). The adverse working environment may lead to massive operating and maintenance costs of turbines. Deploying unmanned aerial vehicles (UAVs) for turbine inspection is considered as a viable alternative to manual inspections. An important objective of automated UAV inspection is to minimize the flight time of the UAVs to inspect all the turbines. A first contribution of this paper is thus formulating an optimization problem to compute the optimal routes for turbine inspection satisfying the above goal. On the other hand, the limited computational capability on UAVs can be used to increase the power generation of wind turbine. Power generation from the turbines can be optimized by controlling the yaw angle of the turbines. Forecasting wind conditions such as wind speed and wind direction is crucial for solving both optimization problems. Therefore, UAVs can utilize their limited computational capability to perform wind forecasting. In this way, UAVs form edge intelligence in offshore wind farm. With the forecasted wind conditions, we design two algorithms to solve the formulated problems, and then evaluate the proposed methods with real-world data. The results reveal that the proposed methods offer an improvement of 44% of the power generation from the turbine compared to hour-ahead forecasting and 25% reduction of the flight time of the UAVs compared to the chosen baseline method.

Given a software defined wide-area network (WAN), how to choose location of controllers and how to assign the controllers to forwarding devices are two significant issues. Previously, most of solutions to these two problems focus on the propagation delay but ignore the balance of controllers, because it's difficult to solve them with consideration of balance of controllers. In this paper, a novel approach which can efficiently and accurately solve SDN controller placement problem and assignment problem for WAN is proposed. The SDN controller assignment problem is formulated as a minimum weight matching of bipartite graph, and it also considers the balance of controllers. The Kuhn- Munkres algorithm based solution is used to find optimal matching between switches and controllers. Then, a genetic algorithm is proposed to solve the controller placement problem based on the controller assignment scheme. The performance shows that our approach has good performance in reducing the average propagation delay between SDN forwarding devices and controllers, and it also achieves better balance of controllers.

Software Defined Networking (SDN) provides flexible and convenient means to support fine-grained management by the logically centralized control. Thus SDN can not only result in better network capacity utilization but can offer better customer satisfaction. In this paper, we analyze and consolidate the utility theory of network bandwidth allocation to offer better customers' satisfaction in SDNs especially when SDNs are incrementally introduced into an existing network. The utilities are modeled as sigmoid curves, since they are well-known functions and often used to describe the perception of Quality of Service (QoS). We propose an optimization bandwidth allocation strategy to maximize the network utility and thus increase the customer satisfaction. The results show that the network utility based on customer satisfaction improvements are possible with proper bandwidth allocation even only a part of SDN forwarding devices in a network topology. Compared with other bandwidth allocation strategies based on fairness, our strategy is more efficient in fulfilling the basic demand of customers.