Papers

Very wideband apertures are needed in positioning, sensing, spectrum monitoring, and modern spread spectrum, e.g., frequency hopping systems. Vivaldi antennas are one of the prominent choices for the aforementioned systems due to their natural wideband characteristics. Furthermore, tightly-coupled antenna arrays have been researched in the recent years to extend the lower band edge of compact arrays by taking advantage of the strong mutual coupling between the elements especially with dipole elements, but not with dual-polarized Vivaldi antennas. This paper presents a novel tightly-coupled dual-polarized antipodal Vivaldi antenna (TC-AVA) with -6 dB impedance bandwidth of 3 to 20 GHz. The tight coupling by overlapping the Vivaldi leaves is shown to extend the lower band edge from 3.75 to 3 GHz and 2.75 GHz, an improvement of 20% to 25% for both polarizations, compared with an isolated antipodal Vivaldi element.

This paper addresses the emulation-based observer design for networked control systems (NCS) with linear plants that operate at two time scales in the presence of measurement noise. The system is formulated as a hybrid singularly perturbed dynamical system, enabling the systematic use of singular perturbation techniques to derive explicit bounds on the maximum allowable transmission intervals (MATI) for both fast and slow communication channels. Under the resulting conditions, the proposed observer guarantees that the estimation error satisfies a global exponential derivative-input-to-state stability (DISS)-like property, where the ultimate bound scales proportionally with the magnitudes of the measurement noise and the time derivative of the control input. The effectiveness of the approach is illustrated through a numerical example.

We address the problem of identifying the dynamical law governing the evolution of a population of indistinguishable particles, when only aggregate distributions at successive times are observed. Assuming a Markovian evolution on a discrete state space, the task reduces to estimating the underlying transition probability matrix from distributional data. We formulate this inverse problem within the framework of entropic optimal transport, as a joint optimization over the transition matrix and the transport plans connecting successive distributions. This formulation results in a convex optimization problem, and we propose an efficient iterative algorithm based on the entropic proximal method. We illustrate the accuracy and convergence of the method in two numerical setups, considering estimation from independent snapshots and estimation from a time series of aggregate observations, respectively.

This paper presents a computationally efficient approach for robust Model Predictive Control of nonlinear systems by combining Random Fourier Features with tube-based MPC. Tube-based Model Predictive Control provides robust constraint satisfaction under bounded model uncertainties arising from approximation errors and external disturbances. The Random Fourier Features method approximates nonlinear system dynamics by solving a numerically tractable least-squares problem, thereby reducing the approximation error. We develop the integration of RFF-based residual learning with tube MPC and demonstrate its application to an autonomous vehicle path-tracking problem using a nonlinear bicycle model. Compared to the linear baseline, the proposed method reduces the tube size by approximately 50%, leading to less conservative behavior and resulting in around 70% smaller errors in the test scenario. Furthermore, the proposed method achieves real-time performance while maintaining provable robustness guarantees.

The state of the art for model predictive control (MPC)-based distributed Q-learning is limited to first-order gradient updates of the MPC parameterization. In general, using secondorder information can significantly improve the speed of convergence for learning, allowing the use of higher learning rates without introducing instability. This work presents a second-order extension to MPC-based Q-learning with updates distributed across local agents, relying only on locally available information and neighbor-to-neighbor communication. In simulation the approach is demonstrated to significantly outperform first-order distributed Q-learning.

In this study, we compare a model reference control (MRC) strategy against conventional PID controllers (tuned via metaheuristic algorithms) for surge velocity control of a thruster-driven marine system, under combined wave disturbance and sensor noise. The goal is to evaluate not only tracking performance but also control energy usage and actuator stress. A high-order identified model of a Blue Robotics T200 thruster with a 2~kg vehicle is used, with an 8~N sinusoidal wave disturbance applied and white noise ( added to the speed measurement. Results show that the optimized MRC (MRC-R*) yields the lowest control energy and smoothest command among all controllers, while maintaining acceptable tracking. The IMC-based design performs closely. In contrast, PID controllers achieve comparable RMS tracking error but at the cost of excessive actuator activity and energy use, making them impractical in such scenarios. Future

Electric vehicles (EVs) have drastically changed the auto industry and developed a new era of technologies where power electronics play the leading role in traction management, energy conversion and vehicle control processes. Nevertheless, this is a digital transformation, and the cyber-attack surface area has increased considerably, to the point that EV traction power electronics are becoming vulnerable to various cybersecurity risks. This paper is able to provide its expertise on possible cyber-attack vectors which can attack important parts of the traction, powertrain, including things like inverters, motor controllers, and communicated systems within the embedded bits. Using the (STRIDE) threat modeling framework, the research outlines and groups the vulnerabilities of the architecture and runs some attack simulations, such as the Denial of Service (DoS), spoofing, firmware manipulation, and data injection. The experiments prove the fact that a slight interruption in the control signal, the sensed data may lead to the severe working implications, such as unstable sensor values of the torque, abnormal voltage shifts, and entire system freezes. These results highlight the high priority on the need of injective embedded intrusion preventive mechanisms and secure design of firmware in EV powertrain electronics. In this paper, the author makes his contribution to the general body of knowledge that underpins the links existing between cyber security practices and the peculiar needs of automotive power electronics.

Large Language Models (LLMs) have demonstrated remarkable potential in hardware front-end design using hardware description languages (HDLs). However, their inherent tendency toward hallucination often introduces functional errors into the generated HDL designs. To address this issue, we propose the framework CorrectHDL that leverages high-level synthesis (HLS) results as functional references to correct potential errors in LLM-generated HDL designs.The input to the proposed framework is a C/C++ program that specifies the target circuit's functionality. The program is provided to an LLM to directly generate an HDL design, whose syntax errors are repaired using a Retrieval-Augmented Generation (RAG) mechanism. The functional correctness of the LLM-generated circuit is iteratively improved by comparing its simulated behavior with an HLS reference design produced by conventional HLS tools, which ensures the functional correctness of the result but can lead to suboptimal area and power efficiency. Experimental results demonstrate that circuits generated by the proposed framework achieve significantly better area and power efficiency than conventional HLS designs and approach the quality of human-engineered circuits. Meanwhile, the correctness of the resulting HDL implementation is maintained, highlighting the effectiveness and potential of agentic HDL design leveraging the generative capabilities of LLMs and the rigor of traditional correctness-driven IC design flows.

Recent advances in Wireless Physical Layer Foundation Models (WPFMs) promise a new paradigm of universal Radio Frequency (RF) representations. However, these models inherit critical limitations found in deep learning such as the lack of explainability, robustness, adaptability, and verifiable compliance with physical and regulatory constraints. In addition, the vision for an AI-native 6G network demands a level of intelligence that is deeply embedded into the systems and is trustworthy. In this vision paper, we argue that the neuro-symbolic paradigm, which integrates data-driven neural networks with rule- and logic-based symbolic reasoning, is essential for bridging this gap. We envision a novel Neuro-Symbolic framework that integrates universal RF embeddings with symbolic knowledge graphs and differentiable logic layers. This hybrid approach enables models to learn from large datasets while reasoning over explicit domain knowledge, enabling trustworthy, generalizable, and efficient wireless AI that can meet the demands of future networks.

Channel state information (CSI)-based user equipment (UE) positioning with neural networks – referred to as neural positioning – is a promising approach for accurate off-device UE localization. Most existing methods train their neural networks with ground-truth position labels obtained from external reference positioning systems, which requires costly hardware and renders label acquisition difficult in large areas. In this work, we propose a novel neural positioning pipeline that avoids the need for any external reference positioning system. Our approach trains the positioning network only using CSI acquired off-device and relative displacement commands executed on commercial off-the-shelf (COTS) robot platforms, such as robotic vacuum cleaners – such an approach enables inexpensive training of accurate neural positioning functions over large areas. We evaluate our method in three real-world scenarios, ranging from small line-of-sight (LoS) areas to larger non-line-of-sight (NLoS) environments, using CSI measurements acquired in IEEE 802.11 Wi-Fi and 5G New Radio (NR) systems. Our experiments demonstrate that the proposed neural positioning pipeline achieves UE localization accuracies close to state-of-the-art methods that require externally acquired high-precision ground-truth position labels for training.

We present VersaPants, the first loose-fitting, textile-based capacitive sensing system for lower-body motion capture, built on the open-hardware VersaSens platform. By integrating conductive textile patches and a compact acquisition unit into a pair of pants, the system reconstructs lower-body pose without compromising comfort. Unlike IMU-based systems that require user-specific fitting or camera-based methods that compromise privacy, our approach operates without fitting adjustments and preserves user privacy. VersaPants is a custom-designed smart garment featuring 6 capacitive channels per leg. We employ a lightweight Transformer-based deep learning model that maps capacitance signals to joint angles, enabling embedded implementation on edge platforms. To test our system, we collected approximately 3.7 hours of motion data from 11 participants performing 16 daily and exercise-based movements. The model achieves a mean per-joint position error (MPJPE) of 11.96 cm and a mean per-joint angle error (MPJAE) of 12.3 degrees across the hip, knee, and ankle joints, indicating the model's ability to generalize to unseen users and movements. A comparative analysis of existing textile-based deep learning architectures reveals that our model achieves competitive reconstruction performance with up to 22 times fewer parameters and 18 times fewer FLOPs, enabling real-time inference at 42 FPS on a commercial smartwatch without quantization. These results position VersaPants as a promising step toward scalable, comfortable, and embedded motion-capture solutions for fitness, healthcare, and wellbeing applications.

A joint communication and computation (JCC) framework using segmented pinching antenna system (PASS) is proposed, where both the communication bit streams and computation data are simultaneously transmitted via uplink communications. The segmented PASS design is used to yield the tractable uplink transmission, and to mitigate large-scale path loss and in-waveguide loss. Based on three operating protocols, namely segment selection (SS), segment aggregation (SA), and segment multiplexing (SM), the joint transmit and receive beamforming problem is formulated: 1) The mean square error (MSE) minimization problem is formulated for computation-oriented cases. To address this problem, a low-complexity alternating optimization-minimum mean square error (AO-MMSE) algorithm is developed. This problem is decomposed into receiver-side and transmitter-side MSE subproblems that are iteratively optimized by MMSE receivers to obtain the closed-form solutions. It is mathematically proved that the segmented JCC-PASS framework significantly outperforms the conventional PASS for the average in-waveguide propagation gain. 2) The weighted sum rate (WSR) maximization problem is formulated for communication-oriented cases. To solve the decomposed receiver-side and transmitter-side MSE subproblems, the AO-weighted minimum mean square error (AO-WMMSE) algorithm is further developed. An auxiliary weight variable is introduced to linearize the WSR function and is alternatively optimized based on WMMSE to derive the closed-form solutions. Simulation results demonstrate that: i) The proposed JCC-PASS framework achieves up to 70.65% and 45.32% reductions in MSE compared with conventional MIMO and conventional PASS, and ii) it reaches 87.70% and 51.35% improvements in WSR compared with conventional MIMO and conventional PASS, respectively.

Optimal operation of chemical processes is vital for energy, resource, and cost savings in chemical engineering. The problem of optimal operation can be tackled with reinforcement learning, but traditional reinforcement learning methods face challenges due to hard constraints related to quality and safety that must be strictly satisfied, and the large amount of required training data. Chemical processes often cannot provide sufficient experimental data, and while detailed dynamic models can be an alternative, their complexity makes it computationally intractable to generate the needed data. Optimal control methods, such as model predictive control, also struggle with the complexity of the underlying dynamic models. Consequently, many chemical processes rely on manually defined operation recipes combined with simple linear controllers, leading to suboptimal performance and limited flexibility. In this work, we propose a novel approach that leverages expert knowledge embedded in operation recipes. By using reinforcement learning to optimize the parameters of these recipes and their underlying linear controllers, we achieve an optimized operation recipe. This method requires significantly less data, handles constraints more effectively, and is more interpretable than traditional reinforcement learning methods due to the structured nature of the recipes. We demonstrate the potential of our approach through simulation results of an industrial batch polymerization reactor, showing that it can approach the performance of optimal controllers while addressing the limitations of existing methods.

Preventive control is a crucial strategy for power system operation against impending natural hazards, and its effectiveness fundamentally relies on the realism of scenario generation. While most existing studies employ sequential Monte Carlo simulation and assume independent sampling of component failures, this oversimplification neglects the spatial correlations induced by meteorological factors such as hurricanes. In this paper, we identify and address the gap in modeling spatial dependence among component failures under extreme weather. We analyze how the mean, variance, and correlation structure of weather intensity random variables influence the correlation of component failures. To fill this gap, we propose a spatially dependent sampling method that enables joint sampling of multiple component failures by generating correlated meteorological intensity random variables. Comparative studies show that our approach captures long-tailed scenarios and reveals more extreme events than conventional methods. Furthermore, we evaluate the impact of scenario selection on preventive control performance. Our key findings are: (1) Strong spatial correlations in uncertain weather intensity consistently lead to interdependent component failures, regardless of mean value level; (2) The proposed method uncovers more high-severity scenarios that are missed by independent sampling; (3) Preventive control requires balancing load curtailment and over-generation costs under different scenario severities; (4) Ignoring failure correlations results in underestimating risk from high-severity events, undermining the robustness of preventive control strategies.

Dynamic graph signal processing provides a principled framework for analyzing time-varying data defined on irregular graph domains. However, existing joint time-vertex transforms such as the joint time-vertex fractional Fourier transform assign only one fractional order to the spatial domain and another one to the temporal domain, thereby restricting their capacity to model the complex and continuously varying dynamics of graph signals. To address this limitation, we propose a novel dynamic multiple-parameter joint time-vertex fractional Fourier transform (DMPJFRFT) framework, which introduces time-varying fractional parameters to achieve adaptive spectral modeling of dynamic graph structures. By assigning distinct fractional orders to each time step, the proposed transform enables dynamic and flexible representation of spatio-temporal signal evolution in the joint time-vertex spectral domain. Theoretical properties of the DMPJFRFT are systematically analyzed, and two filtering approaches: a gradient descent-based method and a neural network-based method, are developed for dynamic signal restoration. Experimental results on dynamic graph and video datasets demonstrate that the proposed framework effectively captures temporal topology variations and achieves superior performance in denoising and deblurring tasks compared with some state-of-the-art graph-based transforms and neural networks.

Parkinson's disease (PD) is a neurodegenerative disorder associated with the accumulation of misfolded alpha-synuclein aggregates, forming Lewy bodies and neuritic shape used for pathology diagnostics. Automatic analysis of immunohistochemistry histopathological images with Deep Learning provides a promising tool for better understanding the spatial organization of these aggregates. In this study, we develop an automated image processing pipeline to segment and classify these aggregates in whole-slide images (WSIs) of midbrain tissue from PD and incidental Lewy Body Disease (iLBD) cases based on weakly supervised segmentation, robust to immunohistochemical labelling variability, with a ResNet50 classifier. Our approach allows to differentiate between major aggregate morphologies, including Lewy bodies and neurites with a balanced accuracy of $80\%$. This framework paves the way for large-scale characterization of the spatial distribution and heterogeneity of alpha-synuclein aggregates in brightfield immunohistochemical tissue, and for investigating their poorly understood relationships with surrounding cells such as microglia and astrocytes.

Rydberg atomic quantum receivers have been seen as novel radio frequency measurements and the high sensitivity to a large range of frequencies makes it attractive for communications reception. However, current implementations of Rydberg array antennas predominantly rely on simple stacking of multiple single-antenna units. While conceptually straightforward, this approach leads to substantial system bulkiness due to the unique requirements of atomic sensors, particularly the need for multiple spatially separated laser setups, rendering such designs both impractical for real-world applications and challenging to fabricate. This limitation underscores the critical need for developing multiplexed Rydberg sensor array architectures. In the domain of conventional RF array antennas, hybrid analog-digital beamforming has emerged as a pivotal architecture for large-scale millimeter-wave (mmWave) multiple-input multiple-output (MIMO) systems, as it substantially reduces the hardware complexity associated with fully-digital beamforming while closely approaching its performance. Drawing inspiration from this methodology, we conduct a systematic study in this work on the design principles, equivalent modeling, and precoding strategies for low-complexity multiplexed Rydberg array, an endeavor crucial to enabling practical and scalable quantum-enhanced communication systems.

Most control systems run on digital hardware with limited communication resources. This work develops self-triggered control for linear systems where sensors update independently (asynchronous measurements). The controller computes an optimal horizon at each sampling instant, selecting which sensor to read over the next several time steps to maximize inter-sample intervals while maintaining stability. Two implementations address computational complexity. The online version solves an optimization problem at each update for theoretical optimality. The offline version precomputes optimal horizons using conic partitioning, reducing online computation to a lookup. Both guarantee exponential stability for unperturbed systems and global uniform ultimate boundedness for systems with bounded disturbances. Simulations demonstrate 59-74% reductions in sensor utilization compared to periodic sampling. The framework enables resource-efficient control in networked systems with communication constraints.

The describing function (DF) and phase response curve (PRC) are classical tools for the analysis of feedback oscillations and rhythmic behaviors, widely used across control engineering, biology, and neuroscience. These tools are known to have limitations in networks of relaxation oscillators and excitable systems. For this reason, the paper proposes a novel approach tailored to excitable systems. Our analysis focuses on the discrete-event operator mapping input trains of events to output trains of events. The methodology is illustrated on the excitability model of Hodgkin-Huxley. The proposed framework provides a basis for designing and analyzing central pattern generators in networks of excitable neurons, with direct relevance to neuromorphic control and neurophysiology.

Model Predictive Control evolved as the state of the art paradigm for safety critical control tasks. Control-as-Inference approaches thereof model the constrained optimization problem as a probabilistic inference problem. The constraints have to be implemented into the inference model. A recently introduced physics-informed Gaussian Process method uses Control-as-Inference with a Gaussian likelihood for state constraint modeling, but lacks guarantees of open-loop constraint satisfaction. We mitigate the lack of guarantees via an additional sampling step using Hamiltonian Monte Carlo sampling in order to obtain safe rollouts of the open-loop dynamics which are then used to obtain an approximation of the truncated normal distribution which has full probability mass in the safe area. We provide formal guarantees of constraint satisfaction while maintaining the ODE structure of the Gaussian Process on a discretized grid. Moreover, we show that we are able to perform optimization of a quadratic cost function by closed form Gaussian Process computations only and introduce the Matérn kernel into the inference model.

Obstructive sleep apnea (OSA) is a highly prevalent sleep disorder that is associated with increased risks of cardiovascular morbidity and all-cause mortality. While existing diagnostic approaches can roughly classify OSA severity or detect isolated respiratory events, they lack the precision and comprehensiveness required for high resolution, event level diagnosis. Here, we present UT OSANet, a deep learning based model designed as a event level, multi scenario diagnostic tool for OSA. This model facilitates detailed identification of events associated with OSA, including apnea, hypopnea, oxygen desaturation, and arousal. Moreover, the model employs flexibly adjustable input modalities such as electroencephalography (EEG), airflow, and SpO 2. It utilizes a random masked modality combination training strategy, allowing it to comprehend cross-modal relationships while sustaining consistent performance across varying modality conditions. This model was trained and evaluated utilizing 9,021 polysomnography (PSG) recordings from five independent datasets. achieving sensitivities up to 0.93 and macro F1 scores of 0.84, 0.85 across home, clinical, and research scenarios. This model serves as an event-level, multi-scenario diagnostic instrument for real-world applications of OSA, while also establishing itself as a means to deepen the mechanistic comprehension of respiratory processes in sleep disorders and their extensive health implications.

Neural audio codecs (NACs) provide compact representations that can be leveraged in many downstream applications, in particular large language models. Yet most NACs encode mixtures of multiple sources in an entangled manner, which may impede efficient downstream processing in applications that need access to only a subset of the sources (e.g., analysis of a particular type of sound, transcription of a given speaker, etc). To address this, we propose a source-aware codec that encodes individual sources directly from mixtures, conditioned on source type prompts. This enables user-driven selection of which source(s) to encode, including separately encoding multiple sources of the same type (e.g., multiple speech signals). Experiments show that our model achieves competitive resynthesis and separation quality relative to a cascade of source separation followed by a conventional NAC, with lower computational cost.

Spectral Graph Neural Networks offer a principled approach to graph filtering but face a fundamental "Stability-vs-Adaptivity" trade-off. This trade-off is dictated by the choice of spectral domain. Filters in the finite [-1, 1] domain (e.g., ChebyNet) are numerically stable at high polynomial degrees (K) but are static and low-pass, causing them to fail on heterophilic graphs. Conversely, filters in the semi-infinite [0, infty) domain (e.g., KrawtchoukNet) are highly adaptive and achieve SOTA results on heterophily by learning non-low-pass responses. However, as we demonstrate, these adaptive filters can also suffer from numerical instability, leading to catastrophic performance collapse at high K. In this paper, we propose to resolve this trade-off by designing a hybrid-domain GNN, HybSpecNet, which combines a stable `ChebyNet` branch with an adaptive `KrawtchoukNet` branch. We first demonstrate that a "naive" hybrid architecture, which fuses the branches via concatenation, successfully unifies performance at low K, achieving strong results on both homophilic and heterophilic benchmarks. However, we then prove that this naive architecture fails the stability test. Our K-ablation experiments show that this architecture catastrophically collapses at K=25, exactly mirroring the collapse of its unstable `KrawtchoukNet` branch. We identify this critical finding as "Instability Poisoning," where `NaN`/`Inf` gradients from the adaptive branch destroy the training of the model. Finally, we propose and validate an advanced architecture that uses "Late Fusion" to completely isolate the gradient pathways. We demonstrate that this successfully solves the instability problem, remaining perfectly stable up to K=30 while retaining its SOTA performance across all graph types. This work identifies a critical architectural pitfall in hybrid GNN design and provides the robust architectural solution.

Accurate and efficient thermal dynamics models of permanent magnet synchronous motors are vital to efficient thermal management strategies. Physics-informed methods combine model-based and data-driven methods, offering greater flexibility than model-based methods and superior explainability compared to data-driven methods. Nonetheless, there are still challenges in balancing real-time performance, estimation accuracy, and explainability. This paper presents a hardware-efficient complex neural dynamics model achieved through the linear decoupling, diagonalization, and reparameterization of the state-space model, introducing a novel paradigm for the physics-informed method that offers high explainability and accuracy in electric motor temperature estimation tasks. We validate this physics-informed method on an NVIDIA A800 GPU using the JAX machine learning framework, parallel prefix sum algorithm, and Compute Unified Device Architecture (CUDA) platform. We demonstrate its superior estimation accuracy and parallelizable hardware acceleration capabilities through experimental evaluation on a real electric motor.

Spectral GNNs, like ChebyNet, are limited by heterophily and over-smoothing due to their static, low-pass filter design. This work investigates the "Adaptive Orthogonal Polynomial Filter" (AOPF) class as a solution. We introduce two models operating in the [-1, 1] domain: 1) `L-JacobiNet`, the adaptive generalization of `ChebyNet` with learnable alpha, beta shape parameters, and 2) `S-JacobiNet`, a novel baseline representing a LayerNorm-stabilized static `ChebyNet`. Our analysis, comparing these models against AOPFs in the [0, infty) domain (e.g., `LaguerreNet`), reveals critical, previously unknown trade-offs. We find that the [0, infty) domain is superior for modeling heterophily, while the [-1, 1] domain (Jacobi) provides superior numerical stability at high K (K>20). Most significantly, we discover that `ChebyNet`'s main flaw is stabilization, not its static nature. Our static `S-JacobiNet` (ChebyNet+LayerNorm) outperforms the adaptive `L-JacobiNet` on 4 out of 5 benchmark datasets, identifying `S-JacobiNet` as a powerful, overlooked baseline and suggesting that adaptation in the [-1, 1] domain can lead to overfitting.

Application of neuromorphic edge devices for control is limited by the constraints on gradient-free online learning and scalability of the hardware across control problems. This paper introduces a synaptic Q-learning algorithm for the control of the classical Cartpole, where the Bellman equations are incorporated at the synaptic level. This formulation enables the iterative evolution of the network topology, represented as a directed graph, throughout the training process. This is followed by a similar approach called neuromorphic Bellman Memory Units (BMU(s)), which are implemented with the Neural Engineering Framework on Intel's Loihi neuromorphic chip. Topology evolution, in conjunction with mixed-signal computation, leverages the optimization of the number of neurons and synapses that could be used to design spike-based reinforcement learning accelerators. The proposed architecture can potentially reduce resource utilization on board, aiding the manufacturing of compact application-specific neuromorphic ICs. Moreover, the on-chip learning introduced in this work and implemented on a neuromorphic chip can enable adaptation to unseen control scenarios.

Joint automatic speech recognition (ASR) and speaker diarization aim to answer the question "who spoke what" in multi-speaker scenarios. In this paper, we present an end-to-end speech large language model (Speech-LLM) for Joint strEamable DIarization and aSr (JEDIS-LLM). The model is trained only on short audio under 20s but is capable of streamable inference on long-form audio without additional training. This is achieved by introducing a Speaker Prompt Cache (SPC) with an on-the-fly update mechanism during chunk-wise streaming inference, inspired by the autoregressive nature of LLMs. The SPC also allows the seamless use of pre-enrolled speaker profiles which is common in many scenarios like meeting transcription. To further enhance diarization capability, we incorporate word-level speaker supervision into the speech encoder during training. Experimental results demonstrate that our system outperforms strong baselines, including Sortformer and Meta-Cat in the local setting on audio up to 20s, and DiarizationLM on long-form audio, despite being fully end-to-end and streamable while DiarizationLM follows a cascaded offline pipeline. To the best of our knowledge, this is the first work enabling zero-shot streamable joint ASR and diarization on long audio using a Speech-LLM trained only on short audio, achieving state-of-the-art performance.

Reconstructing complete traffic flow time-space diagrams from vehicle trajectories offer a comprehensive view on traffic dynamics at arterial intersections. However, obtaining full trajectories across networks is costly, and accurately inferring lane-changing (LC) and car-following behaviors in multi-lane environments remains challenging. This study proposes a generative framework for arterial vehicle trajectory reconstruction that jointly models lane-changing and car-following behaviors through physics-informed multi-task joint learning. The framework consists of a Lane-Change Generative Adversarial Network (LC-GAN) and a Trajectory-GAN. The LC-GAN models stochastic LC behavior from historical trajectories while considering physical conditions of arterial intersections, such as signal control, geometric configuration, and interactions with surrounding vehicles. The Trajectory-GAN then incorporates LC information from the LC-GAN with initial trajectories generated from physics-based car-following models, refining them in a data-driven manner to adapt to dynamic traffic conditions. The proposed framework is designed to reconstruct complete trajectories from only a small subset of connected vehicle (CV) trajectories; for example, even a single observed trajectory per lane, by incorporating partial trajectory information into the generative process. A multi-task joint learning facilitates synergistic interaction between the LC-GAN and Trajectory-GAN, allowing each component to serves as both auxiliary supervision and a physical condition for the other. Validation using two real-world trajectory datasets demonstrates that the framework outperforms conventional benchmark models in reconstructing complete time-space diagrams for multi-lane arterial intersections. This research advances the integration of trajectory-based sensing from CVs with physics-informed deep learning.

Joint admission control and power minimization are critical challenges in intelligent reflecting surface (IRS)-assisted networks. Traditional methods often rely on \( l_1 \)-norm approximations and alternating optimization (AO) techniques, which suffer from high computational complexity and lack robust convergence guarantees. To address these limitations, we propose a sigmoid-based approximation of the \( l_0 \)-norm AC indicator, enabling a more efficient and tractable reformulation of the problem. Additionally, we introduce a penalty dual decomposition (PDD) algorithm to jointly optimize beamforming and admission control, ensuring convergence to a stationary solution. This approach reduces computational complexity and supports distributed implementation. Moreover, it outperforms existing methods by achieving lower power consumption, accommodating more users, and reducing computational time.

Formal guarantees for cyber-physical systems (CPS) rely on diverse assumptions. If satisfied, these assumptions enable the transfer of abstract guarantees into real-world assurances about the deployed CPS. Although assumptions are central to assured CPS, there is little systematic knowledge about what assumptions are made, what guarantees they support, and what it would take to specify them precisely. To fill this gap, we present a survey of assumptions and guarantees in the control, verification, and runtime assurance areas of CPS literature. From 104 papers over a 10-year span (2014-2024), we extracted 423 assumptions and 321 guarantees using grounded-theory coding. We also annotated the assumptions with 21 tags indicating elementary language features needed for specifications. Our analysis highlighted prevalent trends and gaps in CPS assumptions, particularly related to initialization, sensing, perception, neural components, and uncertainty. Our observations culminated in a call to action on reporting and testing CPS assumptions.