Fixed broadband operators are increasingly challenged by the growing complexity of fiber access networks, heterogeneous home Wi-Fi environments, rising customer expectations, and the need to improve operational efficiency while maintaining service quality. Conventional service assurance approaches largely rely on rule-based monitoring, fragmented operational support systems, and manual intervention by Network Operations Center (NOC) teams. Although Artificial Intelligence (AI) and Machine Learning (ML) have been applied to fault prediction, anomaly detection, and customer experience analytics, these implementations often operate as isolated solutions with limited autonomy and cross-domain coordination.Recent advances in Agentic Artificial Intelligence (Agentic AI) provide an opportunity to transform service assurance through intelligent agents capable of reasoning, planning, knowledge retrieval, and workflow execution. This paper proposes a conceptual framework for integrating Agentic AI into fixed broadband service assurance processes. The framework combines operational data sources, predictive ML models, Large Language Model (LLM)-based intelligence, multi-agent orchestration, and human-in-theloop governance to support proactive and explainable operational decision-making.The proposed architecture demonstrates how Agentic AI can enhance key assurance functions, including automated root cause analysis, proactive Quality of Experience (QoE) degradation detection, intelligent incident triage, and NOC copilot assistance. Unlike existing studies that focus on standalone AI applications or mobile network scenarios, this work specifically addresses the operational realities of fixed broadband environments. The findings suggest that Agentic AI can reduce Mean Time to Repair (MTTR), improve customer experience, enhance decision consistency, and increase operational efficiency while maintaining governance and regulatory compliance. The study provides both theoretical insights and practical guidance for telecom operators progressing toward intelligent and autonomous service assurance.

The substantial growth of Internet video streaming has greatly increased resource utilization, particularly bandwidth and energy consumption, especially in systems secured through watermarking. However, the issue of decarbonization has received limited attention. As Quality of Experience (QoE) becomes a central focus, the complexity of these challenges intensifies. Adaptive bitrate algorithms often prioritize video quality while neglecting the associated energy costs, frequently selecting the highest available bitrate and thereby increasing energy consumption. Achieving a balance between QoE and energy efficiency requires careful optimization across multiple factors, including server-side processing, network stability, and bandwidth utilization. While watermarking enhances content protection and ensures secure delivery, it introduces additional computational overhead that increases energy consumption. This study aims to achieve seamless, high-quality video streaming by managing the trade-off between streaming security and QoE while minimizing energy impact. The findings underscore the importance of balancing robust security mechanisms with energy-aware optimization to achieve sustainable, secure, and high-quality video streaming. Numerical results obtained from real-world streaming scenarios validate the efficiency of the proposed method. Even with a 10% increase in the number of connected users, the proposed system reduces origin egress traffic by 25%. At the same time, CDN traffic increases by 23% due to the delivery of higher-bitrate video streams, indicating a 13% improvement in perceived video quality rather than a mere redistribution of system load.

General service readiness in health facilities often fails to predict the quality of care for sick child health services. This study aims to describe the specific service readiness in 8 low- and middle-income countries (LMICs) and examine its relationship with quality of care and user experience. Data was drawn from the Service Provision Assessment surveys, comprehensive national health system assessments conducted within the last decade across Afghanistan, the Democratic Republic of the Congo, Ethiopia, Haiti, Malawi, Nepal, Senegal, and Tanzania. Specific service readiness was identified using the World Health Organization's Service Availability and Readiness Assessment, calculated as the mean preparedness across facilities. Quality of care was assessed by the number of actions in international guidelines that health providers performed. User experience was measured by reported client issues during visits. Multilevel linear regression models were performed testing the relationship between specific service readiness and quality of care and user experience. A total of 5,311 facilities and 20,880 sick child visits were analyzed. The mean score for specific service readiness was 0.57, with the staff and guidelines being the scarcest at 0.34. Health providers completed an average of 5.77 out of 20 recommended actions, indicating suboptimal quality. Interestingly, clients reported a higher user experience. Specific service readiness was positively related to the quality of care, but showed no correlation with user experience. Specific service readiness shows a notable association with the clinical quality of sick child health services in LMICs. Key strategies include staff training, improved availability of equipment and medications. Addressing these prerequisites may be associated with better health outcomes in sick child health services. Not applicable.

Background: A physically active lifestyle is crucial for long-term cardiovascular health; however, access to supervised exercise programs for children and adolescents with congenital heart disease (CHD) remains limited. Although prior digital exercise interventions for this population have demonstrated safety and feasibility, adherence has often been low. Mobile health approaches integrating gamification may enhance motivation and engagement, particularly among young “digital natives.” FOOTLOOSE is an app-based home exercise program developed specifically for children and adolescents with CHD. This study aimed to evaluate user experience, usability, and perceived impact using a multimethod approach. Methods: Children and adolescents aged 10–18 years with simple, moderate, or complex CHD were recruited between July and December 2025 mainly during routine outpatient visits at the TUM Klinikum Deutsches Herzzentrum. Participants used the FOOTLOOSE app in their daily lives over a two-week period. Evaluation included semi-structured qualitative interviews and standardized questionnaires assessing physical activity self-efficacy, enjoyment of physical activity (PACES-S), user experience (UEQ), and health-related quality of life (KINDL®). Interviews were conducted digitally, transcribed verbatim, and analyzed using qualitative content analysis according to Kuckartz until thematic saturation was reached. Results: A total of 22 participants (mean age 13.4 ± 2.3 years; 54.5% female) were included. Overall, the FOOTLOOSE app was perceived positively, with participants highlighting enjoyment, intuitive usability, and personalized workout creation. Participants contributed diverse and creative suggestions for further app development, particularly regarding more advanced gamification features (e.g., games or rankings). Most participants reported self-perceived increase in physical activity during the intervention period (n = 15). UEQ scores (mean ± SD) were as follows: attractiveness (1.3 ± 0.8), perspicuity (1.7 ± 1.1), efficiency (1.2 ± 0.9), dependability (1.4 ± 0.7), stimulation (1.0 ± 1.1), and novelty (0.6 ± 1.0). Conclusions: This study demonstrates the feasibility and user acceptance of a gamified, app-based home exercise program for children and adolescents with CHD. User-centered feedback highlights important directions for iterative refinement, particularly regarding age-appropriate and engaging gamification elements. These findings provide a foundation for future studies evaluating long-term engagement and effectiveness in larger samples.

Networked virtual reality (NVR) whiteboards are increasingly important for enabling geographically dispersed users to engage in real-time idea sharing, collaborative design, and discussion. However, latency caused by network limitations, rendering delays, or synchronization issues can significantly degrade the Quality of Experience (QoE) in whiteboard collaboration. To systematically investigate the impact of latency, this study classified QoE into pragmatic and hedonic aspects, each comprising multiple sub-dimensions. Controlled experiments were conducted to identify the sub-dimensions most affected by latency, which were then adopted as the primary QoE indicators, with the aim of uncovering the processes and mechanisms through which latency shapes QoE. Building on this, we further examined how these impacts vary across different collaboration modes, namely sequential collaboration (SC) for structured design workflows and free collaboration (FC) for open discussion. We also compared two VR whiteboard types, one with avatars (VR+) and the other without avatars (VR), and included a traditional PC-based whiteboard as a baseline. This multidimensional design enables a comprehensive evaluation of latency's impact on QoE across collaboration modes and platforms, providing practical guidance for optimizing NVR whiteboard systems under real-world network and system constraints.

The use of beyond-real interactions (BRI) expands the possibilities of virtual reality (VR) experiences by enabling novel capabilities, such as extending the range of manipulation or employing unusual body configurations. However, such interactions require users to adapt to novel sensorimotor mappings, which can lead to decreased performance and a poorer quality of user experience. Recently, proprioceptive suppression with noisy tendon electrical stimulation (n-TES) has been proposed to promote adaptation to BRI, though its effects have been reported to be limited. This study investigated the effects of proprioceptive suppression via n-TES on the adaptation process to BRI and the consequent changes in body schema, which have remained unexplored in prior research. Using a between-subjects design with 24 participants, we examined the effects of n-TES on a reaching task with Go-Go interaction, which extends virtual hands to enable users to reach farther than physically possible. Body schema changes were measured using a pointing task without visual feedback of the hand after adaptation to the Go-Go interaction. The results revealed that n-TES significantly altered the adaptation process by facilitating early learning and also enhanced body schema changes and aftereffects of adaptation under certain conditions. However, parameter fitting of the learning curves also suggested that n-TES might lead to a deterioration in the final level of performance achieved through adaptation. Our findings provide initial evidence that proprioceptive suppression by n-TES can contribute to facilitating sensorimotor adaptation within VR applications, while its practical utility may require adjusting stimulation to prevent potential side effects in the later stages of adaptation, specifically in trajectory smoothness.

The transition 4G to 5G and beyond in the era of mobile communication has posed unprecedented challenges on scalability, energy efficiency (EE) and sustainability within the Radio Access Network (RAN) segment.In this paper, we propose a renewable-aware convex optimization model and develop a framework applicable to Open Radio Access Network (ORAN) architectures addressing joint Virtual machines (VM) placement, bandwidth allocation, and power control under hybrid grid-renewable energy environments.The proposed model includes an explicit renewable smoothing mechanism in order to overcome the fluctuation of renewable power supply, all the while guaranteeing a good Quality of Service (QoS) and Quality of Experience (QoE) provisioning.Obtaining the problem as a convex optimization ensures global optimality and computational feasibility in large-scale applications too.A formal rounding-and-repair algorithm is proposed to convert relaxed solutions into deployable discrete schedules with bounded integrality gap.Extensive simulations show that the approach can realize a well-balanced resource provisioning between grid and renewables, which in turn provides substantial EE improvement and smooth powerprofile transformation among time slots.Comprehensive comparisons against (Mixed-Integer Problem) MILP, greedy heuristic, and deep reinforcement learning (DRL) baselines demonstrate that the proposed approach achieves within 2.1% of optimal while being 47 faster than MILP.A comparison between the solver types shows that SQP gives the fastest convergence with fewest iterations and same good solution for real-time ORAN optimization.Results illustrate that the smoothing weight parameter () is an appropriate policy control knob, providing a direct balance between renewable stability and total operational cost.Systematic parameter sweeps and stress tests validate robustness under varying network scales, realistic renewable traces, and binding advanced constraints including interference management and QoE provisioning.In summary, the research lays an in-depth groundwork for long-term and sustainable ORAN systems with renewable energy supply to support future carbon-neutral 6G varieties with smart power management and adaptive resource optimization.

Studies on in-vehicle conversational agents have relied on pre-scripted prompts, constraining driver–agent interaction. The present study investigates an LLM-based in-vehicle agent capable of driver-initiated, multi-turn dialogues under manual driving conditions. Forty drivers participated in a driving simulator study using a within-subjects design comparing three conditions: no agent, pre-scripted agent, and ChatGPT-based agent. Results showed that the ChatGPT-based agent condition was associated with more stable driving performance across multiple metrics, including lower variability in longitudinal acceleration, lateral acceleration, and lane deviation. Subjective evaluations indicated higher competence, animacy, affective trust, and overall preference relative to the pre-scripted agent. Thematic analysis revealed diverse conversational patterns beyond driving tasks, includingassistance requests and anthropomorphic interactions. While using a Wizard-of-Oz setup, the findings suggest meaningful benefits for user experience and interaction quality in challenging situations; it suggests that generative, driver-initiated dialogue may support stable driving behavior and enriched user experience in emotionally challenging scenarios.

Abstract Video-based Cognitive radio networks (CRNs) are a sub-type in which some video users send video traffic. In video CRN context, cognitive users must find an optimal rate/power assignment strategy by solving an optimization problem to maximize their perceived quality of experience (QoE) under interference and energy constraints. Due to large dimension of parameter space in dense CRNs, solving this problem using traditional methods such as game theory or other analytical gradient descent-based approaches can lead to large computational burden. Basically, in such scenarios, each cognitive video user actually faces with mean interference effect from its surrounding nodes and must adopt its behavior (rate/power optimization) strategy based on this mean-field interference effect. Because of inherent nature of mean-field game (MFG) theory in distributed solving of high-dimensional optimization problems, it seems to be a good solution candidate in this context. So, in the current paper, we have used MFG for green (energy-efficient) quality control of cognitive users in dense CRNs. We design two different solution approaches based on a finite-difference method (named GMFQ) and machine learning (named D2GMFQ). The first approach is a standard MFG solution but lacks good scalability in very dense CRN scenarios. So, we introduce the second methodology which is fast enough to tackle such cases. Numerical results show that the proposed methods, outperforms similar ones in maximizing sum perceived cognitive user QoEs under energy-efficiency constraints. Specifically, it is determined that about 11 dB and 15 dB gains can be achieved by GMFQ in average comparing with the traditional TCP and UDP streaming scenarios respectively.

Chronic conditions (CC) require continuous, coordinated primary health care (PHC), yet Argentina's health system, including Mendoza province, remains fragmented across public, social security, and private subsectors. Evidence on how people with CC (PwCC) experience this fragmented system is limited. This study compared individuals with and without CC using the People's Voice Survey (PVS) in Mendoza. Cross-sectional, population-based telephone survey in Mendoza province (September-November 2022). From a random sample of 30 000 phone numbers, 1188 adults completed the survey. Weighted analyses compared PwCC and those without (non-PwCC) across domains of healthcare access, use, quality of care, user experience, and confidence in the system. Of respondents, 41% reported ≥1 CC. PwCC had higher healthcare utilization (92.8% vs. 82.2% with ≥1 visit in the past year; P < .001) and were more likely to have a usual source of care (87.0% vs. 80.8%; P = .008), with a greater proportion attached to secondary-level facilities (42% vs. 35%; P < .05). Yet unmet need was also more prevalent among PwCC (22.8% vs. 17.5%; P = .046), driven primarily by long wait times. PwCC rated interpersonal dimensions of their most recent visit more favourably: clear communication (60.5% vs. 52.2%; P = .013), staff dedication (57.2% vs. 49.2%; P = .017), and kindness (56.9% vs. 49.8%; P = .035); yet more frequently reported a perceived medical error (84.3% vs 90.4% reporting no error; P = .003). Mental health needs were more prevalent among PwCC (17.5% vs. 11.6%; P = .007); more PwCC received mental healthcare in the past year (20.2% vs. 10.6%; P < .001), but the proportion with unmet mental health needs was similar across groups (P = .83). Cancer screening rates were similarly low across groups. Confidence in obtaining and affording good-quality care was low across both groups (31%-33%). Together, these patterns reveal a contact-quality-access paradox: greater system engagement and better-rated interpersonal care among PwCC coexist with greater unmet need, more perceived medical errors, and no advantage in financial confidence. Improving health system performance for PwCC requires addressing structural and interpersonal dimensions of care. Strengthening PHC as the foundation for CC management, integrating preventive and mental health services, and building mechanisms for accountability and public input are priority actions.

This study presents a machine learning-based framework for video quality assessment (VQA). This framework enables the mapping of objective metric outputs, such as Structural Similarity Index Measure (SSIM) and Video Multimethod Assessment Fusion (VMAF), to subjective Mean Opinion Score (MOS). The absence of a standardized conversion from objective VQA methods to MOS poses challenges for consistency and comparability across different evaluation methods. The proposed mapping function delivers real-time estimation of user-perceived video quality, offering an effective alternative to traditional subjective tests, which are time-consuming, costly, and unsuitable for online services such as IPTV or streaming platforms. All constructed models were trained using the XGBoost algorithm. The dataset comprises over 700 distorted video sequences, spanning the two most prevalent codecs (H.264 and H.265), along with a wide range of resolutions, bitrates, and packet loss rates. Although real-world network traffic is affected by various impairments, including delay and jitter, packet loss is the primary factor driving video quality degradation. The SSIM-to-MOS and VMAF-to-MOS regression models achieved a Pearson correlation of 0.95 and an RMSE of 0.31. Additionally, we propose a reduced-reference classifier that estimates video quality using a limited set of video characteristics, including a frame (packet) loss rate derived from the original sequence. This approach enables the model to evaluate video quality without requiring a direct comparison between the original and the test sequence. The model achieves a weighted F1 score of 0.92 and is suitable for deployment in time-sensitive services such as live streaming and IPTV, even in environments where packet loss may occur. Benchmarking against a Back Propagation Neural Network (BPNN) and state-of-the-art methods confirmed the superiority of the described approach. All source codes are publicly available to support reproducibility and further research in Quality of Experience (QoE) assessment.

It is well-known that, without access to application-layer parameters controlled by Over-The-Top (OTT) providers, Mobile Network Operators (MNOs) struggle to accurately predict customers’ Quality of Experience (QoE). While some previous proposals have suggested interaction between OTTs and MNOs, they have faced challenges in terms of practical implementation and limited application scenarios. This work aims to advance these solutions with two key contributions. First, following the Open Radio Access Network (O-RAN) architecture, we propose adding components that integrate a machine learning (ML)-based QoE prediction model, deployed by the MNO, into the O-RAN system. By establishing specific data-sharing interfaces between OTTs and MNOs, our approach helps MNOs overcome the limitations in updating their quality prediction modules. Second, we present a network-aware, ML-driven QoE prediction model that captures the relationship between the resulting QoE and various network parameters, such as signal-to-interference-noise ratio (SINR), channel quality indicator (CQI), network resource blocks (RBs), throughput, and device mobility. Among seven considered ML regressors, the Gradient Boosting (GB) achieved the highest QoE prediction performance in terms of R 2 (0.906) and RMSE (0.259).

Mobile Edge Computing (MEC) shifts the computing power to the edge of core networks and provides important impetus in the flourishment of delay sensitive services at the network edge. Digital Twin (DT) technique enables object behavior monitoring, analysis, and prediction through data analytics and artificial intelligence, which facilitates inference service provisioning based on machine learning models. In this paper, we deal with the Quality-of-Experience (QoE) issue of user satisfaction on inference services in DT-assisted MEC networks, through executing user tasks locally or offloaded to the MEC network. We formulate two novel optimization problems: the utility maximization problem, and the dynamic utility maximization problem, with the aim to maximize the total utility of user task executions in terms of QoEs and service delays of users with the services. We first provide an Integer Linear Programming solution for the utility maximization problem when the problem size is small or medium; otherwise we devise a randomized algorithm with high probability, at the expense of bounded resource violations. We then develop an efficient online heuristic for the dynamic utility maximization problem. We also devise an online algorithm with a provable competitive ratio for a special case of the dynamic utility maximization problem without the bandwidth constraint. We finally evaluate the performance of proposed algorithms through simulations. The simulation results show that the proposed algorithms are promising.

With the introduction of highly automated vehicles (HAV; SAE Level 4) driverless mobility systems may fundamentally change public transportation. For potential passengers of these systems, new situations may arise, leaving them without a driver to directly communicate with. This can be problematic when the driving automation is confronted with situations it might be unable to manage by itself, which would cause a minimal risk maneuver (MRM), e.g. a complete standstill of the vehicle. Some concepts enhance these systems with a remote assistant who can support the HAV, adding novelty to the transportation process. This increased novelty may result in user discomfort for passengers, causing a need for information systems that address this. One possibility to do so and to thus improve hedonic quality and thus comfort for passengers is by providing system transparency via a media rich in-vehicle Human-Machine Interface (iHMI), which provides information to passengers about the ADS and its processes. However, there is still a gap in research on the ideal communication mechanism to ensure good hedonic quality and comfort during these situations. We conducted a simulator study in virtual reality (VR), investigating interfaces based on media richness theory to address this issue. The interface provided a multi-layered iHMI consisting of transparency information presented with varying levels of media richness. In a block design, participants experienced three versions of an iHMI, one presenting information via text, another combining text and auditive presentation and a third adding a human avatar simulating Face-to-Face communication. After each scenario, participants completed questionnaires regarding understanding, predictability, trust and user experience. Study results reveal significant increases in the hedonic quality of user experience and trust towards the system for interfaces with increased media richness, providing a first step towards enhancing user comfort with iHMI in automated mobility systems during challenging situations for the automation system, such as MRM.

As new applications evolve rapidly, wireless networks increasingly require low-delay communication to significantly enhance the quality of user experience. In response, the evolution of the medium access control (MAC) layer has gained more attention, particularly through the application of reinforcement learning to optimize access strategies. In order to meet the low-delay requirements, we propose a reinforcement learning-based MAC protocol, named soft actor-critic multiple access (SAC-MA). To mitigate frequent collisions caused by the exploratory behavior, we propose a multiple waiting actions mechanism that allows stations to wait for multiple time slots. This mechanism enables the agent to develop a more flexible and intelligent access strategy, thereby effectively reducing delay. Additionally, we introduce an innovative formulation in which the head-of-line packet is treated as the agent, enabling more timely feedback and observations. We conduct extensive simulations to demonstrate that SAC-MA: 1) reduces delay by approximately 27.9% and 56.5% compared to the conventional MAC protocol with standard parameters under the collision and capture models, respectively; 2) adapts to environmental changes in dynamic scenarios; 3) coexists harmoniously with legacy stations and reduces the network delay in heterogeneous scenarios. Finally, we perform ablation studies to evaluate the effectiveness of the proposed mechanisms.

Real-time video streaming over wireless networks has become increasingly reliant on adaptive bitrate (ABR) control to mitigate variability in bandwidth, latency, and user mobility. However, existing ABR algorithms are predominantly reactive, operate on limited network observability, and largely ignore the computational and bandwidth overhead introduced by encryption, which is now ubiquitous in edge-assisted multimedia delivery. This paper presents a machine-learning driven adaptive bitrate optimization framework that jointly addresses predictive bandwidth estimation, mobility dynamics, and security constraints in edge-assisted video transmission. We formulate bitrate selection as a stochastic optimization problem and develop a cross-layer system model that characterizes network evolution, user mobility, and cryptographic overhead. An edge-hosted learning engine leverages supervised prediction and reinforcement-driven control to proactively select bitrates using features derived from transport behavior, playback state, and security cost. We implement the proposed approach in a prototype edge-streaming platform and evaluate performance under realistic wireless traces, user mobility patterns, and multi-user contention. Experimental results demonstrate that the proposed system reduces stall probability by up to 42%, improves average Quality of Experience (QoE) by up to 27%, and maintains equitable performance under multi-user load, while introducing only modest cryptographic overhead. We further analyze the security–performance trade-offs, identify risk factors in encrypted media pipelines, and quantify the operational limits of edge execution. The results highlight the importance of integrating prediction, security-awareness, and scalability into ABR design, and demonstrate the potential of edge-hosted learning models to enable secure, high-quality, and resource-efficient video streaming in mobile environments.

Meet Indonesia is a provider of high-quality goalkeeper gloves that currently markets its products through social media and e-commerce. However, the owner wants to improve the brand image through a more flexible independent platform to face intense industrial competition. This research aims to design a UI/UX-based sales website using CMS Wordpress to enhance Meet Indonesia's brand image and systematically evaluate its quality. The type of research used is action research involving 15 respondents as samples (2 computer experts, 2 marketing experts, 1 owner, and 10 consumers). Data collection techniques were carried out through observation, interviews, documentation, and Likert scale questionnaires. Data analysis was performed using the WebQual Index method to measure service quality and the HEART Framework to measure user experience. The results of the analysis using WebQual showed a WebQual Index (WQI) value of 91.67%, which falls into the "Very Good" category, while the evaluation using the HEART Framework produced a total overall score of 93.54%, reflecting a very high quality of user experience in the dimensions of Happiness, Engagement, Adoption, Retention, and Task Success. The developed website is proven to be able to provide a significant contribution in improving Meet Indonesia's brand image by presenting a positive experience for users both emotionally and functionally, thereby increasing user trust in the brand and encouraging consumer loyalty

This study explores how video quality, field of view, and latency influence users’ performance, depth perception, and overall experience in remote-control systems for industrial teleoperated applications. In this controlled laboratory investigation, we conducted an experimental study using a test setup that closely replicated real-world conditions while ensuring experimental control and safety. Participants completed 18 trials under varying fields of view (Wide vs. Narrow), latency (High, Medium, Low), and video quality (High, Medium, Low) conditions to evaluate their navigation accuracy, depth perception and Quality of Experience. Using a mixed method approach, considering Quality of Experience and User Experience insights, we integrated objective performance metrics, quantitative user evaluations, and qualitative feedback to understand how visual constraints impact control and decision-making. The results show that latency had the most impact on navigation accuracy, followed by field of view. The high latency level significantly influenced performance, causing larger deviations from the intended path and reducing confidence in estimating depth. The effect was even more pronounced in narrow fields of view, where high latency further amplified navigation difficulties. In contrast, a wider field of view helped mitigate latency effects by improving situational awareness and reducing reliance on secondary visual cues. While video quality had a minimal impact on objective navigation performance, qualitative feedback indicated that higher video quality improved confidence in depth perception and reduced visual strain, supporting users in making precise judgments. We also collected users’ feedback that provided cues to further enhance the teleoperation systems.

ABSTRACT In the field of multimedia, conventional video streaming remains the dominant playback format. Current research predominantly focuses on optimizing adaptive bitrate (ABR) algorithms to enhance quality of experience (QoE), delivering improved viewing experiences to users. However, the majority of existing approaches consider only single‐user scenarios, whereas practical environments necessitate addressing the challenge of multiple users sharing bottleneck link bandwidth. These methods fail to holistically consider the multiple factors influencing QoE and provide insufficient consideration for QoE fairness. While bandwidth allocation fairness is achieved, ensuring fairness in user QoE remains challenging. Furthermore, these ABR algorithms rely solely on bitrate for adaptation, resulting in limited control dimensions and an inability to provide fine‐grained ABR decisions. Additionally, certain methods require the deployment of additional control equipment to obtain global network states for achieving fairness, which increases deployment complexity in existing networks. To address the issue of QoE fairness in multi‐user video streaming, this paper models it as a Markov decision process (MDP) for multi‐agent cooperative fair allocation of limited bottleneck link resources, and proposes a multi‐agent reinforcement learning‐based ABR algorithm. The algorithm incorporates several improved Multi‐Agent Deep Reinforcement Learning (MADRL) techniques to collaboratively select optimal chunk bitrate and download delay for different users, allowing for fine‐grained control of the chunk download strategies. Furthermore, this paper designs and implements a video streaming distribution framework that operates without relying on additional network‐assisted devices. This framework can efficiently acquire global client state, overcome performance disparities among clients, and achieve centralized and scalable ABR decision‐making. Experimental results demonstrate that compared to existing methods, the proposed approach achieves a significant rightward shift in the CDF curve of average user QoE at the 50th percentile. Furthermore, it adeptly selects appropriate bitrate strategies for different types of devices. Consequently, the total transmitted data volume is reduced by 30.4% to 54.3%, leading to optimized bandwidth occupancy while ensuring user QoE fairness.

Serverless edge computing (SEC) that integrates serverless and edge computing paradigms has facilitated the deployment of intelligent Internet-of-things (IoT) applications. In SEC systems, energy efficiency and serverless pricing are essential to maintain operational sustainability. Nevertheless, most existing energy-saving techniques focus only on stable energy scenarios and are therefore inapplicable to energy-harvesting SEC systems powered by intermittent renewable sources. On the other hand, serverless pricing policies generally neglect the personalized perceptions of user quality-of-experience (QoE) preferences, thereby resulting in holistic user QoE degradation from a system perspective. Moreover, these approaches cannot guarantee functional correctness of serverless applications due to the appearance of computation and communication errors in practical SEC systems. To tackle these challenges, we investigate the preference-aware fault-tolerant function embedding problem for enhancing the holistic user QoE in energy-harvesting SEC systems. We first design a personalized QoE preference predictor to characterize trade-offs between service completion time and resultant service fees of individual users. Subsequently, we develop a reinforcement learning method to decide static function embedding decisions at the offline phase. Considering the intermittency of renewable sources, we further provide an energy-adaptive function replica freezing strategy at the online phase. Evaluations demonstrate that our approach boosts the holistic user QoE by 32.2% over state-of-the-art algorithms.