Papers

Passive acoustic monitoring has become a key strategy in biodiversity assessment, conservation, and behavioral ecology, especially as Internet-of-Things (IoT) devices enable continuous in situ audio collection at scale. While recent self-supervised learning (SSL)-based audio encoders, such as BEATs and AVES, have shown strong performance in bioacoustic tasks, their computational cost and limited robustness to unseen environments hinder deployment on resource-constrained platforms. In this work, we introduce BioME, a resource-efficient audio encoder designed for bioacoustic applications. BioME is trained via layer-to-layer distillation from a high-capacity teacher model, enabling strong representational transfer while reducing the parameter count by 75%. To further improve ecological generalization, the model is pretrained on multi-domain data spanning speech, environmental sounds, and animal vocalizations. A key contribution is the integration of modulation-aware acoustic features via FiLM conditioning, injecting a DSP-inspired inductive bias that enhances feature disentanglement in low-capacity regimes. Across multiple bioacoustic tasks, BioME matches or surpasses the performance of larger models, including its teacher, while being suitable for resource-constrained IoT deployments. For reproducibility, code and pretrained checkpoints are publicly available.

Future 6G non-terrestrial networks aim to deliver ubiquitous connectivity to remote and undeserved regions, but unmanned aerial vehicle (UAV) base stations face fundamental challenges such as limited numbers and power budgets. To overcome these obstacles, high-altitude platform station (HAPS) equipped with a reconfigurable intelligent surface (RIS), so-called HAPS-RIS, is a promising candidate. We propose a novel unified joint multi-objective framework where UAVs and HAPS-RIS are fully integrated to extend coverage and enhance network performance. This joint multi-objective design maximizes the number of users served by the HAPS-RIS, minimizes the number of UAVs deployed and minimizes the total average UAV path loss subject to quality-of-service (QoS) and resource constraints. We propose a novel low-complexity solution strategy by proving the equivalence between minimizing the total average UAV path loss upper bound and k-means clustering, deriving a practical closed-form RIS phase-shift design, and introducing a mapping technique that collapses the combinatorial assignments into a zone radius and a bandwidth-portioning factor. Then, we propose a dynamic Pareto optimization technique to solve the transformed optimization problem. Extensive simulation results demonstrate that the proposed framework adapts seamlessly across operating regimes. A HAPS-RIS-only setup achieves full coverage at low data rates, but UAV assistance becomes indispensable as rate demands increase. By tuning a single bandwidth portioning factor, the model recovers UAV-only, HAPS-RIS-only and equal bandwidth portioning baselines within one formulation and consistently surpasses them across diverse rate requirements. The simulations also quantify a tangible trade-off between RIS scale and UAV deployment, enabling designers to trade increased RIS elements for fewer UAVs as service demands evolve.

This chapter is motivated by the need for a rigorous and comprehensive analysis of the Doppler effects encountered by electromagnetic and acoustic signals across a diverse spectrum of modern applications. These include land mobile communications, various Internet of Things (IoT) networks, machine-type communications (MTC), and various radar and satellite-based systems for navigation and sensing, as well as the emerging regime of integrated sensing and communications (ISAC). A wide array of kinematic profiles is investigated, ranging from uniform motion and constant acceleration to more complex general motion. Consequently, the multi-faceted factors influencing the Doppler shift are addressed in detail, encompassing classical kinematics, special and general relativity, atmospheric dynamics, and the properties of the propagation medium. This work is intended to establish a definitive theoretical foundation for both the general enthusiast and the specialized researcher seeking to master the complexities of signal frequency shifts in modern wireless sensing and communications systems.

Feedback optimization refers to a class of methods that steer a control system to a steady state that solves an optimization problem. Despite tremendous progress on the topic, an important problem remains open: enforcing state constraints at all times. The difficulty in addressing it lies on mediating between the safety enforcement and the closed-loop stability, and ensuring the equivalence between closed-loop equilibria and the optimization problem's critical points. In this work, we present a feedback-optimization method that enforces state constraints at all times employing high-order control-barrier functions. We provide several results on the proposed controller dynamics, including well-posedness, safety guarantees, equivalence between equilibria and critical points, and local and global (in certain convex cases) asymptotic stability of optima. Various simulations illustrate our results.

Applying Nearest Convex Hull Classification (NCHC) to blind user identification in a massive Multiple Input Multiple Output (MIMO) communications system is proposed. The method is blind in the way that the Base Station (BS) only requires a training sequence containing unknown data symbols obtained from the user without further knowledge on the channel, modulation, coding or even noise power. We evaluate the algorithm under the assumption of gaussian transmit signals using the non-rigorous replica method. To facilitate the computations the existence of an Operator Valued Free Fourier Transform is postulated, which is verified by Monte Carlo simulation. The replica computations are conducted in the large but finite system by applying saddle-point integration with inverse temperature $\beta$ as the large parameter. The classifier accuracy is estimated by gaussian approximation through moment-matching.

This tutorial-style overview article examines the fundamental principles and methods of robustness, using wireless sensing and communication (WSC) as the narrative and exemplifying framework. First, we formalize the conceptual and mathematical foundations of robustness, highlighting the interpretations and relations across robust statistics, optimization, and machine learning. Key techniques, such as robust estimation and testing, distributionally robust optimization, and regularized and adversary training, are investigated. Together, the costs of robustness in system design, for example, the compromised nominal performances and the extra computational burdens, are discussed. Second, we review recent robust signal processing solutions for WSC that address model mismatch, data scarcity, adversarial perturbation, and distributional shift. Specific applications include robust ranging-based localization, modality sensing, channel estimation, receive combining, waveform design, and federated learning. Through this effort, we aim to introduce the classical developments and recent advances in robustness theory to the general signal processing community, exemplifying how robust statistical, optimization, and machine learning approaches can address the uncertainties inherent in WSC systems.

In this work, we present Covo-Audio, a 7B-parameter end-to-end LALM that directly processes continuous audio inputs and generates audio outputs within a single unified architecture. Through large-scale curated pretraining and targeted post-training, Covo-Audio achieves state-of-the-art or competitive performance among models of comparable scale across a broad spectrum of tasks, including speech-text modeling, spoken dialogue, speech understanding, audio understanding, and full-duplex voice interaction. Extensive evaluations demonstrate that the pretrained foundation model exhibits strong speech-text comprehension and semantic reasoning capabilities on multiple benchmarks, outperforming representative open-source models of comparable scale. Furthermore, Covo-Audio-Chat, the dialogue-oriented variant, demonstrates strong spoken conversational abilities, including understanding, contextual reasoning, instruction following, and generating contextually appropriate and empathetic responses, validating its applicability to real-world conversational assistant scenarios. Covo-Audio-Chat-FD, the evolved full-duplex model, achieves substantially superior performance on both spoken dialogue capabilities and full-duplex interaction behaviors, demonstrating its competence in practical robustness. To mitigate the high cost of deploying end-to-end LALMs for natural conversational systems, we propose an intelligence-speaker decoupling strategy that separates dialogue intelligence from voice rendering, enabling flexible voice customization with minimal text-to-speech (TTS) data while preserving dialogue performance. Overall, our results highlight the strong potential of 7B-scale models to integrate sophisticated audio intelligence with high-level semantic reasoning, and suggest a scalable path toward more capable and versatile LALMs.

This paper presents a low-jitter hybrid voltage level shifter (HVLS) architecture for high-speed mixed-signal and digital power-amplifier applications. The proposed design employs a regenerative cross-coupled feedback network to simultaneously generate two synchronized voltage-domain outputs (e.g., the nominal supply voltage VDDL and one at twice its value, VDDH = 2VDDL). This enables direct drive of cascoded class-D power amplifiers without additional delay-calibration circuitry, amongst other mixed-signal applications. A prototype HVLS circuit, together with an impedance-matching network and a pre-driver for high-speed off-chip characterization, was fabricated in a 22-nm FD-SOI process technology. The total die area, including all interface circuitry, is 477 x 462 um^2, while the active area of the HVLS core is 3.26 x 2 um^2. Measured at 12.2 GHz, the circuit dissipates 427.8 uW and achieves a measured rms jitter below 43 fs (10 Hz-500 kHz), limited by the noise floor of the measurement setup.

Manual annotation of the images of thin tissue sections remains a time-consuming step in Mueller microscopy and limits its scalability. We present a novel automated approach using only the total intensity M11 element of the Mueller matrix as an input to a U-Net architecture with a pretrained ResNet-34 encoder. The network was trained to distinguish four classes in the images of murine uterine cervix sections: background, internal os, cervical tissue, and vaginal wall. With only 70 cervical tissue sections, the model achieved 89.71% pixel accuracy and 80.96% mean tissue Dice coefficient on the held-out test dataset. Transfer learning from ImageNet enables accurate segmentation despite limited size of training dataset typical of specialized biomedical imaging. This intensity-based framework requires minimal preprocessing and is readily extensible to other imaging modalities and tissue types, with publicly available graphical annotation tools for practical deployment.

Covert communications, also referred to as low probability of detection (LPD) communications, provide a higher level of privacy protection than cryptography and physical-layer security (PLS) by hiding transmissions in the ambient environment. In this work, we investigate covert communications in the presence of a disco reconfigurable intelligent surface (DRIS) deployed by the warden Willie, which reduces Willie's detection error probability (DEP), i.e., the sum of the false alarm rate (FAR) and the miss detection rate (MDR), and degrades the communication performance between Alice and Bob, without relying on either channel state information (CSI) or additional jamming power. However, the introduction of the DRIS makes it analytically intractable for Willie to construct the Neyman-Pearson (NP) detector, which is the optimal detector for monitoring potential covert transmissions between Alice and Bob. To this end, we develop an unsupervised semi-parametric plug-in likelihood-ratio detector for Willie. The proposed detector retains the parametric Gamma reference model under the silent hypothesis without requiring prior knowledge of noise, and learns from unlabeled data a one-dimensional monotone normalizing flow model for the analytically intractable distribution under the transmission hypothesis. In particular, it exploits the structural prior inherent in covert communications that Willie's observations reduce to noise only when Alice and Bob are silent. The monitoring performance at Willie is evaluated in terms of DEP, while the communication impact on Alice and Bob is quantified by the signal-to-jamming-plus-noise ratio (SJNR). Simulation results verify the analysis and show that the proposed unsupervised plug-in likelihood-ratio detector achieves monitoring performance close to that of its supervised counterpart.

Ensuring robust security in near-field Integrated Sensing and Communication (ISAC) systems remains a critical challenge due to dynamic channel conditions, multi-eavesdropper threats, and the high computational burden of real-time optimization at mmWave and THz frequencies. To address these challenges, this paper introduces a novel Bayesian-Stackelberg framework that jointly optimizes sensing, beamforming, and communication. The dual-algorithm design integrates (i) Adaptive Hybrid Node Role Switching between secure transmission and cooperative jamming (ii) Belief-Driven Sensing and Beamforming for confidence based resource allocation. The proposed unified framework significantly improves robustness against attacks while preserving linear computational complexity. Simulation results across carrier frequencies ranging from 28 to 410 GHz demonstrate that the method achieves up to a 35% increase in secrecy rates and a success rate exceeding 98%, outperforming conventional communication systems with minimal runtime overhead. These findings underscore the scalability of belief-driven ISAC security solutions for low-complexity deployment in next generation communications.

We present an open-source implementation of a closed-loop Brain-Computer Interface (BCI) system based on electrocorticographic (ECoG) recordings. Our setup integrates FieldTrip for interfacing with a Micromed acquisition system and PsychoPy for implementing experiments. We open-source three custom Python libraries (psychopylib, pymarkerlib, and pyfieldtriplib) each covering different aspects of a closed-loop BCI interface: designing interactive experiments, sending event information, and real-time signal processing. Our modules facilitate the design and operation of a transparent BCI system, promoting customization and flexibility in BCI research, and lowering the barrier for researchers to translate advances in ECoG decoding into BCI applications.

Rolling element bearings are critical components in rotating machinery, and their condition significantly influences system performance, reliability, and operational lifespan. Timely and accurate fault detection is essential to prevent unexpected failures and reduce maintenance costs. Traditional diagnostic methods often rely on manual feature extraction and shallow classifiers, which may be inadequate for capturing the complex patterns embedded in raw vibration signals. In this study, a compact one-dimensional convolutional neural network (1D CNN) is developed for automated bearing fault diagnosis using raw time-domain vibration data, eliminating the need for manual feature engineering. The model is trained and evaluated on two established benchmark datasets: the Case Western Reserve University (CWRU) dataset and the Paderborn University (PU) dataset. The CWRU data were segmented based on four distinct motor load conditions (0 HP to 3 HP), with each load scenario trained and tested independently to ensure strict separation and prevent data leakage. The CNN achieved high average test accuracies of 99.14%, 98.85%, 97.42%, and 95.14% for 0 HP, 1 HP, 2 HP, and 3 HP, respectively. On the PU dataset, known for its naturally induced faults and greater operational variability the model achieved a robust average testing accuracy of 95.63%. These results affirm the model ability to generalize across datasets and varying operating conditions. Further improvements were observed through hyperparameter tuning, particularly window length and training epochs, underscoring the importance of tailored configurations for specific datasets and load conditions. Overall, the proposed method demonstrates the effectiveness and scalability of 1D CNNs for real-time, data-driven bearing fault diagnosis, offering a reliable foundation for condition monitoring in industrial applications.

Modern applications, such as orchestrating the collective behavior of robotic swarms or traffic flows, require the coordination of large groups of agents evolving in unstructured environments, where disturbances and unmodeled dynamics are unavoidable. In this work, we develop a scalable macroscopic density control framework in which a feedback law is designed directly at the level of an advection–diffusion partial differential equation. We formulate the control problem in the density space and prove global exponential convergence towards the desired behavior in $\mathcal{L}^2$ with guaranteed asymptotic rejection of bounded unknown drift terms, explicitly accounting for heterogeneous agent dynamics, unmodeled behaviors, and environmental perturbations. Our theoretical findings are corroborated by numerical experiments spanning heterogeneous oscillators, traffic systems, and swarm robotics in partially unknown environments.

Beamforming (BF) is essential for enhancing system capacity in fifth generation (5G) and beyond wireless networks, yet exhaustive beam training in ultra-massive multiple-input multiple-output (MIMO) systems incurs substantial overhead. To address this challenge, we propose a deep learning based framework that leverages position-aware features to improve beam prediction accuracy while reducing training costs. The proposed approach uses spatial coordinate labels to supervise a position extraction branch and integrates the resulting representations with beam-domain features through a feature fusion module. A dual-branch RegNet architecture is adopted to jointly learn location related and communication features for beam prediction. Two fusion strategies, namely adaptive fusion and adversarial fusion, are introduced to enable efficient feature integration. The proposed framework is evaluated on datasets generated by the DeepMIMO simulator across four urban scenarios at 3.5 GHz following 3GPP specifications, where both reference signal received power and user equipment location information are available. Simulation results under both in-distribution and out-of-distribution settings demonstrate that the proposed approach consistently outperforms traditional baselines and achieves more accurate and robust beam prediction by effectively incorporating positioning information.

Urban energy systems face increasing challenges due to high penetration of renewable energy sources, extreme weather events, and other high-impact, low-probability disruptions. This project proposes a community-centered, open-access framework to enhance the resilience and reliability of urban power and gas networks by integrating microgrid partitioning, mobile energy storage deployment, and data-driven risk assessment. The approach involves converting passive distribution networks into active, self-healing microgrids using distributed energy resources and remotely controlled switches to enable flexible reconfiguration during normal and emergency operations. To address uncertainties from intermittent renewable generation and variable load, an adjustable interval optimization method combined with a column and constraint generation algorithm is developed, providing robust planning solutions without requiring probabilistic information. Additionally, a real-time online risk assessment tool is proposed, leveraging 25 multi-dimensional indices including load, grid status, resilient resources, emergency response, and meteorological factors to support operational decision-making during extreme events. The framework also optimizes the long-term sizing and allocation of mobile energy storage units while incorporating urban traffic data for effective routing during emergencies. Finally, a novel time-dependent resilience and reliability index is introduced to quantify system performance under diverse operating conditions. The proposed methodology aims to enable resilient, efficient, and adaptable urban energy networks capable of withstanding high-impact disruptions while maximizing operational and economic benefits.

We study the impact of CO 2 absorption on hollow-core fibre transmission. Using spectral pre-equalisation, we digitally post-compensate gas-line absorption and show a 5.5 dB reduction in Q-factor penalty, outperforming a 383-tap equaliser by 1.3 dB.

Integrating domain knowledge into neural networks is a central challenge in scientific machine learning. Three paradigms have emerged – data-driven (Neural Ordinary Differential Equations, NODEs), soft-constrained (Physics-Informed Neural Networks, PINNs), and hard-constrained (Differentiable Programming, DP) – each encoding physical knowledge at different levels of structural commitment. However, how these strategies impact not only predictive accuracy but also downstream tasks such as control synthesis remains insufficiently understood. This paper presents a comparative study of NODEs, PINNs, and DP for dynamical system modeling, using the Single Machine Infinite Bus power system as a benchmark. We evaluate these paradigms across three tasks: trajectory prediction, parameter identification, and Linear Quadratic Regulator control synthesis. Our results yield three principal findings. First, knowledge representation determines generalization: NODE, which learns the system operator, enables robust extrapolation, whereas PINN, which approximates a solution map, restricts generalization to the training horizon. Second, hard-constrained formulations (DP) reduce learning to a low-dimensional physical parameter space, achieving faster and more reliable convergence than soft-constrained approaches. Third, knowledge fidelity propagates to control performance: DP produces controllers that closely match those obtained from true system parameters, while NODE provides a viable data-driven alternative by recovering control-relevant Jacobians with $3-4\%$ relative error and yielding LQR gains within $0.36\%$ of the ground truth. Based on these findings, we propose a practical decision framework for selecting knowledge integration strategies in neural modeling of dynamical systems.

Deterministic frequency deviations (DFDs) are systematic and predictable excursions of grid frequency that arise from synchronized generation ramps induced by electricity market scheduling. In this paper, we analyze the impact of the European day-ahead market reform of 1 October 2025, which replaced hourly trading blocks with quarter-hourly blocks, on DFDs in the Central European synchronous area. Using publicly available frequency measurements, we compare periods before and after the reform based on daily frequency profiles, indicators characterizing frequency deviations, principal component analysis, Fourier-based functional data analysis, and power spectral density analysis. We show that the reform substantially reduces characteristic hourly frequency deviations and suppresses dominant spectral components at hourly and half-hourly time scales, while quarter-hourly structures gain relative importance. While the likelihood of large frequency deviations decreases overall, reductions for extreme events are less clear and depend on the metric used. Our results demonstrate that market design reforms can effectively mitigate systematic frequency deviations, but also highlight that complementary technical and regulatory measures are required to further reduce large frequency excursions in low-inertia power systems.

The agricultural industry has been searching for a high-speed successor to the 250~kbit/s CAN bus backbone of ISO~11783 (ISOBUS) for over a decade, yet no protocol-level solution has reached standardization. Meanwhile, modern planters, sprayers, and Virtual Terminals are already constrained by the bus bandwidth. This paper presents ISO FastLane, a gateway-less dual-stack approach that routes point-to-point ISOBUS traffic over Ethernet while keeping broadcast messages on the existing CAN bus. The solution requires no new state machines, no middleware, and no changes to application layer code: only a simple Layer~3 routing decision and a lightweight peer discovery mechanism called Augmented Address Claim (AACL). Legacy devices continue to operate unmodified and unaware of FastLane traffic. Preliminary tests reported on the paper demonstrate that ISO FastLane accelerates Virtual Terminal object pool uploads by factor of 8 and sustains Task Controller message rates over 100 times beyond the current specification limit. Because ISO FastLane builds entirely on existing J1939 and ISO~11783 conventions, it can be implemented by ISOBUS engineers in a matter of weeks. This is delivering tangible performance gains today, without waiting for the long-term High Speed ISOBUS solution.

Artificial intelligence (AI) has become a key enabler for next-generation wireless communication systems, offering powerful tools to cope with the increasing complexity, dynamics, and heterogeneity of modern wireless environments. To illustrate the role and impact of AI in wireless communications, this paper takes collaborative spectrum sensing (CSS) in cognitive and intelligent wireless networks as a representative application and surveys recent advances from an AI perspective. We first introduce the fundamentals of CSS, including the general framework, classical detector design, fusion strategies and evaluation metrics. Then, we present an overview of the state-of-the-art research on AI-driven CSS, classified into three categories according to learning paradigms: discriminative deep learning (DL), generative DL models, and deep reinforcement learning (DRL). Building on this, we explore AI-empowered semantic communication (SemCom) as a paradigm-shifting solution for CSS. By extracting and transmitting task-relevant features, SemCom upgrades CSS from a computation-centric approach to a highly efficient joint communication and computation framework. Both single-user and multi-user SemCom scenarios are elaborated in detail. Finally, we discuss limitations, open challenges, and future research directions at the intersection of AI and wireless communication.

The Teacher Assignment Problem is a combinatorial optimization problem that involves assigning teachers to courses while guaranteeing that all courses are covered, teachers do not teach too few or too many hours, teachers do not switch assigned courses too often and possibly teach the courses they favor. Typically the problem is solved manually, a task that requires several hours every year. In this work we present a mathematical formulation for the problem and an experimental evaluation of the model implemented using state-of-the-art SMT, CP, and MILP solvers. The implementations are tested over a real-world dataset provided by the Division of Systems and Control at Chalmers University of Technology, and produce teacher assignments with smaller workload deviation, a more even workload distribution among the teachers, and a lower number of switched courses.

Correctly detecting radar targets is usually challenged by clutter and waveform distortion. An additional difficulty stems from the relative proximity of several targets, the latter being perceived as a single target in the worst case, or influencing each other's detection thresholds. The negative impact of targets proximity notably depends on the range resolution defined by the radar parameters and the adaptive threshold adopted. This paper addresses the matter of targets detection in radar range profiles containing multiple targets with varying proximity and distorted echoes. Inspired by recent contributions in the radar and signal processing literature, this work proposes partially complex-valued neural networks as an adaptive range profile processing. Simulated datasets are generated and experiments are conducted to compare a common pulse compression approach with a simple neural network partially defined by complex-valued parameters. Whereas the pulse compression processes one pulse length at a time, the neural network put forward is a generative architecture going through the entire received signal in one go to generate a complete detection profile.

Acoustic room modes admit closed-form expressions for rectangular rooms with perfectly reflecting walls, from which the Green's function can be computed directly through the eigenfunction expansion. First-order approximations also exist for nearly rigid boundaries; however, current analytical methods fail to accommodate more general boundary conditions, e.g., when wall absorption is significant. In this work, we present a comprehensive analysis that extends previous studies by including additional first-order asymptotics that account for soft-wall boundaries. In addition, we introduce a semi-analytical, efficient, and reliable method for computing the Green's function in rectangular rooms, which is described and verified through numerical tests. The resulting error decreases rapidly for sufficiently large truncation order, making the method suitable as a benchmark for numerical simulations. Additional aspects regarding the spectral basis orthogonality and completeness are also addressed, providing a general framework for the validity of the proposed approach.

The joint communications and sensing (JCAS) paradigm is envisioned as a core capability of sixth-generation (6G) wireless networks, enabling the integration of data communication and environmental sensing within a unified system. By reusing spectrum, waveforms, and hardware resources, JCAS improves spectral efficiency, reduces system complexity, and hardware cost, while enabling new use cases. Nevertheless, the realization of JCAS is hindered by inherent trade-offs between communication and sensing objectives, limited controllability of wireless propagation, and stringent hardware and design constraints. Simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RIS) have recently emerged as a promising technology to address these challenges by enabling full-space programmable manipulation of electromagnetic waves. This survey provides a systematic and in-depth review of STAR-RIS-enabled JCAS systems. Specifically, we first introduce the fundamental principles of JCAS and STAR-RIS. We then classify and review the state-of-the-art research on STAR-RIS-assisted JCAS from multiple perspectives, encompassing system architectures, waveform and beamforming design, resource allocation, optimization frameworks, and learning-based control. Finally, we identify key open challenges that remain unsolved and outline promising future research directions toward intelligent, flexible, and perceptive 6G wireless networks.

Unmanned Aerial Vehicles (UAVs) are crucial for advancing railway communication by offering reliable connectivity, adaptive coverage, and mobile edge services . This survey examines UAV-assisted approaches for 6G railway needs including ultra-reliable low-latency communication (URLLC) and integrated sensing and communication (ISAC). We cover railway channel models, reconfigurable intelligent surfaces (RIS), and UAV-assisted mobile edge computing (MEC). Key challenges include coexistence with existing systems, handover management, Doppler effect, and security. The roadmap suggests work on integrated communication-control systems and AI-driven optimization for intelligent railway networks.

Low-latency live streaming (LLS) has emerged as a popular web application, with many platforms adopting real-time protocols such as WebRTC to minimize end-to-end latency. However, we observe a counter-intuitive phenomenon: even when the actual encoded bitrate does not fully utilize the available bandwidth, stalling events remain frequent. This insufficient bandwidth utilization arises from the intrinsic temporal variations of real-time video encoding, which cause conventional packet-level congestion control algorithms to misestimate available bandwidth. When a high-bitrate frame is suddenly produced, sending at the wrong rate can either trigger packet loss or increase queueing delay, resulting in playback stalls. To address these issues, we present Camel, a novel frame-level congestion control algorithm (CCA) tailored for LLS. Our insight is to use frame-level network feedback to capture the true network capacity, immune to the irregular sending pattern caused by encoding. Camel comprises three key modules: the Bandwidth and Delay Estimator and the Congestion Detector, which jointly determine the average sending rate, and the Bursting Length Controller, which governs the emission pattern to prevent packet loss. We evaluate Camel on both large-scale real-world deployments and controlled simulations. In the real-world platform with 250M users and 2B sessions across 150+ countries, Camel achieves up to a 70.8% increase in 1080P resolution ratio, a 14.4% increase in media bitrate, and up to a 14.1% reduction in stalling ratio. In simulations under undershooting, shallow buffers, and network jitter, Camel outperforms existing congestion control algorithms, with up to 19.8% higher bitrate, 93.0% lower stalling ratio, and 23.9% improvement in bandwidth estimation accuracy.

The interconnection of 5G and non-terrestrial networks (NTNs) has been actively studied to expand connectivity beyond conventional terrestrial infrastructure. In the 3GPP standardization of 5G systems, the 5G Quality of Service (QoS) Identifier (5QI) is defined to characterize the QoS requirements of different traffic requirements. However, it falls short in capturing the diverse latency, capacity, and reliability profiles of NTN environments, particularly when NTNs are used as backhaul. Furthermore, it is difficult to manage individual traffic flows and perform efficient resource allocation and routing when a large number of 5G traffic flows are present in NTN systems. To address these challenges, we propose an optimization framework that enhances QoS handling by introducing an NTN QoS Identifier (NQI) and grouping 5G traffic into NTN slices based on similar requirements. This enables unified resource control and routing for a large number of 5G flows in NTN systems. In this paper, we present the detailed procedure of the proposed framework, which consists of 5QI to NQI mapping, NTN traffic to NTN slice mapping, and slice-level flow and routing optimization. We evaluate the framework by comparing multiple mapping schemes through numerical simulations and analyze their impact on overall network performance.

Preloading is widely used in short video platforms to minimize playback stalls by downloading future content in advance. However, existing strategies face a tradeoff. Aggressive preloading reduces stalls but wastes bandwidth, while conservative strategies save data but increase the risk of playback stalls. This paper presents PromptPream, a computation powered preloading paradigm that breaks this tradeoff by using local computation to reduce bandwidth demand. Instead of transmitting pixel level video chunks, PromptPream sends compact semantic prompts that are decoded into high quality frames using generative models such as Stable Diffusion. We propose three core techniques to enable this paradigm: (1) a gradient based prompt inversion method that compresses frames into small sets of compact token embeddings; (2) a computation aware scheduling strategy that jointly optimizes network and compute resource usage; and (3) a scalable searching algorithm that addresses the enlarged scheduling space introduced by scheduler. Evaluations show that PromptStream reduces both stalls and bandwidth waste by over 31%, and improves Quality of Experience (QoE) by 45%, compared to traditional strategies.

L7 load balancers are a fundamental building block in microservices as they enable fine-grained traffic distribution. Compared to monolithic applications, microservices demand higher performance and stricter isolation from load balancers. This is due to the increased number of instances, longer service chains, and the necessity for co-location with services on the same host. Traditional sidecar-based load balancers are ill-equipped to meet these demands, often resulting in significant performance degradation. In this work, we present XLB, a novel architecture that reshapes L7 load balancers as in-kernel interposition operating on the socket layer. We leverage eBPF to implement the core load balancing logic in the kernel, and address the connection management and state maintenance challenges through novel socket layer redirection and nested eBPF maps designs. XLB eliminates the extra overhead of scheduling, communication, and data movement, resulting in a more lightweight, scalable, and efficient L7 load balancer architecture. Compared to the widely used microservices load balancers (Istio and Cilium), over 50 microservice instances, XLB achieves up to 1.5x higher throughput and 60% lower end-to-end latency.