Hybrid AC/DC transmission grids incorporate Modular Multilevel Converters functioning as Interconnecting Power Converters (IPCs). The control role assigned to each converter significantly influences grid dynamics. Traditionally, these converters operate with static control roles, but recent studies have proposed scheduling their roles based on day-ahead forecasts to enhance stability performance. However, in systems with high renewable energy penetration, forecast deviations can render scheduled control assignments suboptimal or even lead to instability. To address this challenge, this work proposes an online scheduling recalculation algorithm that dynamically adapts IPC control roles during system operation. The approach leverages a data-driven multi-criteria decision-making framework, integrating surrogate models of conventional small-signal stability analysis tools to enable a fast computation of system stability and stability performance indicators.
In this paper, we address the problem of scheduling sensing and communication functionality in an integrated sensing and communication (ISAC) enabled base station (BS) operating in an indoor factory (InF) environment. The BS is performing the task of detecting an AGV while managing downlink transmission of ultra-reliable low-latency communication (URLLC) data in a time-sharing manner. Scheduling fixed time slots for both sensing and communication is inefficient for the InF environment, as the instantaneous environmental changes necessitate a higher frequency of sensing operations to accurately detect the AGV. To address this issue, we propose an environment-aware scheduling scheme, in which we first formulate an optimization problem to maximize the probability of detection of AGV while considering the survival time constraint of URLLC data. Subsequently, utilizing the Nash bargaining theory, we propose an adaptive time-sharing scheme that assigns sensing duration in accordance with the environmental clutter density and distributes time to URLLC depending on the incoming traffic rate. Using our own Python-based discrete-event link-level simulator, we demonstrate the effectiveness of our proposed scheme over the baseline scheme in terms of probability of detection and downlink latency.
To address the power system hardening problem, traditional approaches often adopt robust optimization (RO) that considers a fixed set of concerned contingencies, regardless of the fact that hardening some components actually renders relevant contingencies impractical. In this paper, we directly adopt a dynamic uncertainty set that explicitly incorporates the impact of hardening decisions on the worst-case contingencies, which leads to a decision-dependent uncertainty (DDU) set. Then, a DDU-based robust-stochastic optimization (DDU-RSO) model is proposed to support the hardening decisions on distribution lines and distributed generators (DGs). Also, the randomness of load variations and available storage levels is considered through stochastic programming (SP) in the innermost level problem. Various corrective measures (e.g., the joint scheduling of DGs and energy storage) are included, coupling with a finite support of stochastic scenarios, for resilience enhancement. To relieve the computation burden of this new hardening formulation, an enhanced customization of parametric column-and-constraint generation (P-C&CG) algorithm is developed. By leveraging the network structural information, the enhancement strategies based on resilience importance indices are designed to improve the convergence performance. Numerical results on 33-bus and 118-bus test distribution networks have demonstrated the effectiveness of DDU-RSO aided hardening scheme. Furthermore, in comparison to existing solution methods, the enhanced P-C&CG has achieved a superior performance by reducing the solution time by a few orders of magnitudes.
The manufacturing industry is undergoing a transformative shift, driven by cutting-edge technologies like 5G, AI, and cloud computing. Despite these advancements, effective system control, which is crucial for optimizing production efficiency, remains a complex challenge due to the intricate, knowledge-dependent nature of manufacturing processes and the reliance on domain-specific expertise. Conventional control methods often demand heavy customization, considerable computational resources, and lack transparency in decision-making. In this work, we investigate the feasibility of using Large Language Models (LLMs), particularly GPT-4, as a straightforward, adaptable solution for controlling manufacturing systems, specifically, mobile robot scheduling. We introduce an LLM-based control framework to assign mobile robots to different machines in robot assisted serial production lines, evaluating its performance in terms of system throughput. Our proposed framework outperforms traditional scheduling approaches such as First-Come-First-Served (FCFS), Shortest Processing Time (SPT), and Longest Processing Time (LPT). While it achieves performance that is on par with state-of-the-art methods like Multi-Agent Reinforcement Learning (MARL), it offers a distinct advantage by delivering comparable throughput without the need for extensive retraining. These results suggest that the proposed LLM-based solution is well-suited for scenarios where technical expertise, computational resources, and financial investment are limited, while decision transparency and system scalability are critical concerns.
Connecting aerial and terrestrial users with a single base station (BS) is increasingly challenging due to the rising number of aerial users like unmanned aerial vehicles (UAVs). Traditional BSs, designed with down-tilted beams, focus mainly on ground users, but massive MIMO (mMIMO) systems can significantly enhance coverage in low-altitude airspace. This paper analyzes how a mMIMO BS serves both aerial and terrestrial users in a 3D spectrum-sharing scheme. Using Semi-orthogonal User Selection (SUS) and random scheduling, we assess the spectral efficiency and performance limits of these systems. Results reveal that mMIMO effectively supports more terrestrial users, influenced by channel characteristics and user scheduling strategies, providing key insights for future 3D aerial-terrestrial networks.
This paper addresses query scheduling for goal-oriented semantic communication in pull-based status update systems. We consider a system where multiple sensing agents (SAs) observe a source characterized by various attributes and provide updates to multiple actuation agents (AAs), which act upon the received information to fulfill their heterogeneous goals at the endpoint. A hub serves as an intermediary, querying the SAs for updates on observed attributes and maintaining a knowledge base, which is then broadcast to the AAs. The AAs leverage the knowledge to perform their actions effectively. To quantify the semantic value of updates, we introduce a grade of effectiveness (GoE) metric. Furthermore, we integrate cumulative perspective theory (CPT) into the long-term effectiveness analysis to account for risk awareness and loss aversion in the system. Leveraging this framework, we compute effect-aware scheduling policies aimed at maximizing the expected discounted sum of CPT-based total GoE provided by the transmitted updates while complying with a given query cost constraint. To achieve this, we propose a model-based solution based on dynamic programming and model-free solutions employing state-of-the-art deep reinforcement learning (DRL) algorithms. Our findings demonstrate that effect-aware scheduling significantly enhances the effectiveness of communicated updates compared to benchmark scheduling methods, particularly in settings with stringent cost constraints where optimal query scheduling is vital for system performance and overall effectiveness.
Queueing systems present many opportunities for applying machine-learning predictions, such as estimated service times, to improve system performance. This integration raises numerous open questions about how predictions can be effectively leveraged to improve scheduling decisions. Recent studies explore queues with predicted service times, typically aiming to minimize job time in the system. We review these works, highlight the effectiveness of predictions, and present open questions on queue performance. We then move to consider an important practical example of using predictions in scheduling, namely Large Language Model (LLM) systems, which presents novel scheduling challenges and highlights the potential for predictions to improve performance. In particular, we consider LLMs performing inference. Inference requests (jobs) in LLM systems are inherently complex; they have variable inference times, dynamic memory footprints that are constrained by key-value (KV) store memory limitations, and multiple possible preemption approaches that affect performance differently. We provide background on the important aspects of scheduling in LLM systems, and introduce new models and open problems that arise from them. We argue that there are significant opportunities for applying insights and analysis from queueing theory to scheduling in LLM systems.
복잡한 생성 사전(generative prior) 아래에서 비용 함수(cost function)를 낮추면서도 타깃 분포의 높은 확률 영역을 유지해야 하는 제로차 최적화(zeroth-order optimization) 문제는, 정밀한 기울기 정보 없이도 전역적으로 유의미한 해를 찾아야 한다는 점에서 매우 도전적입니다. 이를 분포 관점으로 바꾸면 목표는 대략 생성 사전 \(L(s)\)와 비용 \(d(s)\)를 결합한 \(L(s)e^{-T\cdot d(s)}\)에 비례하는 분포에서 샘플링하는 문제로 정리되며, 여기서 온도 \(T\)는 탐색과 최적화의 균형을 조절합니다. 저자들은 기존의 모델 기반 최적화(model-based optimization, MBO)가 표현력이 큰 근사 학습기에서는 유한 표본 보장을 제공하기 어렵다는 한계를 지적하고, 이를 보완하기 위해 coarse learnability라는 새로운 통계적 가정을 제안합니다. 이 가정의 핵심은 학습된 모델이 타깃 분포를 정확히 복원하지 못하더라도, 타깃의 확률 질량(probability mass)을 다항식 수준의 손실로 넓게 덮기만 하면 충분하다는 점에 있습니다. 이러한 관점은 생성 모델을 정교한 복원기라기보다, 타깃이 존재하는 영역을 안정적으로 포괄하는 느슨한 커버 기계로 해석하게 하며, 실제 생성 사전의 불완전성을 전제로 한 보다 현실적인 최적화 틀을 제시합니다.
이 가정 위에서 제안된 반복적 MBO 알고리즘 ALDrIFT는 각 단계에서 단순히 학습 분포를 사용하는 대신 sample correction 단계를 통해 타깃 분포의 질량을 보존하도록 보정함으로써, 유한 표본 환경에서도 안정적인 근사를 가능하게 만듭니다. 그 결과 coarse learnability가 성립하는 한, 타깃 분포를 따라가는 데 필요한 샘플 수를 다항식 범위로 억제할 수 있으며, 이는 기존 접근보다 훨씬 약한 전제 위에서 작동하는 점이 중요한 기여입니다. 더 나아가 이 틀은 \(R^n\)에서 quadratic envelope로 상계되는 비볼록 목적함수의 전역 최적화에까지 확장되며, optimistic posterior 계열의 분포가 자연스럽게 좋은 해 영역에 더 많은 질량을 배분하도록 설계될 수 있음을 보입니다. 이때 전역 \(\varepsilon\)-최적성에 도달하는 샘플 복잡도가 \(\widetilde{O}(\log 1/\varepsilon)\)로 정리된다는 점은, 매우 강한 샘플 효율성을 의미하며 optimistic space-partitioning 방법과 유사한 로그형 수렴 성질을 보여줍니다. 또한 저자들은 단순한 가우시안 혼합 예제에서 최대우도추정(maximum likelihood estimation, MLE)과 과도하게 평활화된 커널 밀도 추정기(kernel density estimator, KDE)가 coarse learnability를 자연스럽게 만족할 수 있음을 이론적으로 제시하여, 이 가정이 추상적 개념에 머물지 않음을 뒷받침합니다. 마지막으로 이러한 관점은 추론 시 정렬(inference-time alignment)과도 연결되며, 단순한 설정에서조차 원시적인 대규모 언어 모델(large language model, LLM)이 제로차 피드백을 통해 더 낮은 비용 영역으로 분포를 이동시킬 수 있다는 정성적 증거를 통해 실용적 확장 가능성까지 보여줍니다. 종합하면, 이 연구는 완벽한 생성 모델이 아니라 전역 질량을 충분히 덮는 거친 학습 가능성만으로도 샘플 효율적인 전역 최적화가 가능하다는 점을 이론적으로 정립한 점에서 의미가 큽니다.
In recent years, Mixture-of-Experts (MoE) has emerged as an effective approach for enhancing the capacity of deep neural network (DNN) with sub-linear computational costs. However, storing all experts on GPUs incurs significant memory overhead, increasing the monetary cost of MoE-based inference. To address this, we propose eMoE, a memory efficient inference system for MoE-based large language models (LLMs) by leveraging our observations from experiment measurements. eMoE reduces memory usage by predicting and loading only the required experts based on recurrent patterns in expert routing. To reduce loading latency while maintaining accuracy, as we found using the same experts for subsequent prompts has minimal impact on perplexity, eMoE invokes the expert predictor every few prompts rather than for each prompt. In addition, it skips predictions for tasks less sensitive to routing accuracy. Finally, it has task-aware scheduling to minimize inference latency by considering Service Level Objectives (SLOs), task-specific output lengths, and expert loading latencies. Experimental results show that compared to existing systems, eMoE reduces memory consumption by up to 80% while maintaining accuracy and reduces inference latency by up to 17%. It also enables processing prompts 40x longer, batches 4.5x larger, and achieves 1.5x higher throughput.
To improve the efficiency of distributed large language model (LLM) inference, various parallelization strategies, such as tensor and pipeline parallelism, have been proposed. However, the distinct computational characteristics inherent in the two stages of LLM inference-prefilling and decoding-render a single static parallelization strategy insufficient for the effective optimization of both stages. In this work, we present Seesaw, an LLM inference engine optimized for throughput-oriented tasks. The key idea behind Seesaw is dynamic model re-sharding, a technique that facilitates the dynamic reconfiguration of parallelization strategies across stages, thereby maximizing throughput at both phases. To mitigate re-sharding overhead and optimize computational efficiency, we employ tiered KV cache buffering and transition-minimizing scheduling. These approaches work synergistically to reduce the overhead caused by frequent stage transitions while ensuring maximum batching efficiency. Our evaluation demonstrates that Seesaw achieves a throughput increase of up to 1.78x (1.36x on average) compared to vLLM, the most widely used state-of-the-art LLM inference engine.
This paper proposes an accelerated consensus-based distributed iterative algorithm for resource allocation and scheduling. The proposed gradient-tracking algorithm introduces an auxiliary variable to add momentum towards the optimal state. We prove that this solution is all-time feasible, implying that the coupling constraint always holds along the algorithm iterative procedure; therefore, the algorithm can be terminated at any time. This is in contrast to the ADMM-based solutions that meet constraint feasibility asymptotically. Further, we show that the proposed algorithm can handle possible link nonlinearity due to logarithmically-quantized data transmission (or any sign-preserving odd sector-bound nonlinear mapping). We prove convergence over uniformly-connected dynamic networks (i.e., a hybrid setup) that may occur in mobile and time-varying multi-agent networks. Further, the latency issue over the network is addressed by proposing delay-tolerant solutions. To our best knowledge, accelerated momentum-based convergence, nonlinear linking, all-time feasibility, uniform network connectivity, and handling (possible) time delays are not altogether addressed in the literature. These contributions make our solution practical in many real-world applications.
MoE (Mixture of Experts) prevails as a neural architecture that can scale modern transformer-based LLMs (Large Language Models) to unprecedented scales. Nevertheless, large MoEs' great demands of computing power, memory capacity and memory bandwidth make scalable serving a fundamental challenge and efficient parallel inference has become a requisite to attain adequate throughput under latency constraints. DeepSpeed-MoE, one state-of-the-art MoE inference framework, adopts a 3D-parallel paradigm including EP (Expert Parallelism), TP (Tensor Parallel) and DP (Data Parallelism). However, our analysis shows DeepSpeed-MoE's inference efficiency is largely bottlenecked by EP, which is implemented with costly all-to-all collectives to route token activation. Our work aims to boost DeepSpeed-MoE by strategically reducing EP's communication overhead with a technique named Speculative MoE. Speculative MoE has two speculative parallelization schemes, speculative token shuffling and speculative expert grouping, which predict outstanding tokens' expert routing paths and pre-schedule tokens and experts across devices to losslessly trim EP's communication volume. Besides DeepSpeed-MoE, we also build Speculative MoE into a prevailing MoE inference engine SGLang. Experiments show Speculative MoE can significantly boost state-of-the-art MoE inference frameworks on fast homogeneous and slow heterogeneous interconnects.
We address the problem of object arrangement and scheduling for sequential 3D printing. Unlike the standard 3D printing, where all objects are printed slice by slice at once, in sequential 3D printing, objects are completed one after other. In the sequential case, it is necessary to ensure that the moving parts of the printer do not collide with previously printed objects. We look at the sequential printing problem from the perspective of combinatorial optimization. We propose to express the problem as a linear arithmetic formula, which is then solved using a solver for satisfiability modulo theories (SMT). However, we do not solve the formula expressing the problem of object arrangement and scheduling directly, but we have proposed a technique inspired by counterexample guided abstraction refinement (CEGAR), which turned out to be a key innovation to efficiency.
Sign Languages are the primary form of communication for Deaf communities across the world. To break the communication barriers between the Deaf and Hard-of-Hearing and the hearing communities, it is imperative to build systems capable of translating the spoken language into sign language and vice versa. Building on insights from previous research, we propose a deep learning model for Sign Language Production (SLP), which to our knowledge is the first attempt on Greek SLP. We tackle this task by utilizing a transformer-based architecture that enables the translation from text input to human pose keypoints, and the opposite. We evaluate the effectiveness of the proposed pipeline on the Greek SL dataset Elementary23, through a series of comparative analyses and ablation studies. Our pipeline's components, which include data-driven gloss generation, training through video to text translation and a scheduling algorithm for teacher forcing - auto-regressive decoding seem to actively enhance the quality of produced SL videos.
Large Language Models (LLMs) encounter challenges in efficiently processing long-text queries, as seen in applications like enterprise document analysis and financial report comprehension. While conventional solutions employ long-context processing or Retrieval-Augmented Generation (RAG), they suffer from prohibitive input expenses or incomplete information. Recent advancements adopt context compression and dynamic retrieval loops, but still sacrifice critical details or incur iterative costs. To address these limitations, we propose OkraLong, a novel framework that flexibly optimizes the entire processing workflow. Unlike prior static or coarse-grained adaptive strategies, OkraLong adopts fine-grained orchestration through three synergistic components: analyzer, organizer and executor. The analyzer characterizes the task states, which guide the organizer in dynamically scheduling the workflow. The executor carries out the execution and generates the final answer. Experimental results demonstrate that OkraLong not only enhances answer accuracy but also achieves cost-effectiveness across a variety of datasets.
Recent advancements in probing Large Language Models (LLMs) have explored their latent potential as personalized travel planning agents, yet existing benchmarks remain limited in real world applicability. Existing datasets, such as TravelPlanner and TravelPlanner+, suffer from semi synthetic data reliance, spatial inconsistencies, and a lack of key travel constraints, making them inadequate for practical itinerary generation. To address these gaps, we introduce TripCraft, a spatiotemporally coherent travel planning dataset that integrates real world constraints, including public transit schedules, event availability, diverse attraction categories, and user personas for enhanced personalization. To evaluate LLM generated plans beyond existing binary validation methods, we propose five continuous evaluation metrics, namely Temporal Meal Score, Temporal Attraction Score, Spatial Score, Ordering Score, and Persona Score which assess itinerary quality across multiple dimensions. Our parameter informed setting significantly enhances meal scheduling, improving the Temporal Meal Score from 61% to 80% in a 7 day scenario. TripCraft establishes a new benchmark for LLM driven personalized travel planning, offering a more realistic, constraint aware framework for itinerary generation. Dataset and Codebase will be made publicly available upon acceptance.
Training vision-language models for image-text alignment typically requires large datasets to achieve robust performance. In low-data scenarios, standard contrastive learning can struggle to align modalities effectively due to overfitting and unstable training dynamics. In this paper, we propose a variance-aware loss scheduling approach that dynamically adjusts the weighting of the contrastive loss based on the statistical variability (uncertainty) in the model's alignment predictions. Using a subset of the Flickr8k image-caption dataset to simulate limited data conditions, we demonstrate that our approach improves image-text retrieval accuracy compared to a fixed-weight baseline. We also compare against other adaptive weighting strategies (using output entropy and cosine similarity spread) and find that variance-aware scheduling provides the best overall trade-off. Qualitatively, our method yields more distinct multimodal embeddings as shown by t-SNE visualizations. Moreover, in a stress test with noise-injected captions and images, the variance-guided loss proves more robust, maintaining higher recall when random perturbations are introduced. These results highlight the benefit of adaptive loss weighting for multimodal alignment in low-data regimes.
This study explores the integration of clearing function (CF)-based release planning into Material Requirements Planning (MRP) systems, with a focus on mitigating the inherent rigidity of MRP in handling variability in production environments. By replacing the backward scheduling step of MRP with a CF-based optimization model, this work investigates its impact on overall costs, lead times, and system performance in multi-item, multi-stage production systems. Computational experiments were conducted on two distinct systems - a simple and a complex production system - under varying demand behaviors and utilization levels. The findings reveal that the CF-based release order planning consistently outperforms traditional MRP in managing cost and variability, particularly under scenarios of higher demand uncertainty and shop load. While MRP demonstrates stability in less complex scenarios, its inability to adapt dynamically to variability leads to higher costs in most conditions. The analysis underscores the potential of CF-based methods to enhance planning efficiency in dynamic production settings, providing a foundation for future research on integrating capacity constraints within the broader MRP framework.
Subset selection is central to many wireless communication problems, including link scheduling, power allocation, and spectrum management. However, these problems are often NP-complete, because of which heuristic algorithms applied to solve these problems struggle with scalability in large-scale settings. To address this, we propose a determinantal point process-based learning (DPPL) framework for efficiently solving general subset selection problems in massive networks. The key idea is to model the optimal subset as a realization of a determinantal point process (DPP), which balances the trade-off between quality (signal strength) and similarity (mutual interference) by enforcing negative correlation in the selection of {\em similar} links (those that create significant mutual interference). However, conventional methods for constructing similarity matrices in DPP impose decomposability and symmetry constraints that often do not hold in practice. To overcome this, we introduce a new method based on the Gershgorin Circle Theorem for constructing valid similarity matrices. The effectiveness of the proposed approach is demonstrated by applying it to two canonical wireless network settings: an ad hoc network in 2D and a cellular network serving drones in 3D. Simulation results show that DPPL selects near-optimal subsets that maximize network sum-rate while significantly reducing computational complexity compared to traditional optimization methods, demonstrating its scalability for large-scale networks.
Nonlinear frequency hopping has emerged as a promising approach for mitigating interference and enhancing range resolution in automotive FMCW radar systems. Achieving an optimal balance between high range-resolution and effective interference mitigation remains challenging, especially without centralized frequency scheduling. This paper presents a game-theoretic framework for interference avoidance, in which each radar operates as an independent player, optimizing its performance through decentralized decision-making. We examine two equilibrium concepts--Nash Equilibrium (NE) and Coarse Correlated Equilibrium (CCE)--as strategies for frequency band allocation, with CCE demonstrating particular effectiveness through regret minimization algorithms. We propose two interference avoidance algorithms: Nash Hopping, a model-based approach, and No-Regret Hopping, a model-free adaptive method. Simulation results indicate that both methods effectively reduce interference and enhance the signal-to-interference-plus-noise ratio (SINR). Notably, No-regret Hopping further optimizes frequency spectrum utilization, achieving improved range resolution compared to Nash Hopping.