Search Results (2,586)

Large language models (LLMs) often tend to hallucinate, especially in domain-specific tasks and tasks that require reasoning. Previously, we introduced SubGraph Retrieval Augmented Generation (SG-RAG) as a novel Graph RAG method for multi-hop question answering. SG-RAG leverages Cypher queries to search a given knowledge graph and retrieve the subgraph necessary to answer the question. The results from our previous work showed the higher performance of our method compared to the traditional Retrieval Augmented Generation (RAG). In this work, we further enhanced SG-RAG by proposing an additional step called Merging and Ordering Triplets (MOT). The new MOT step seeks to decrease the redundancy in the retrieved triplets by applying hierarchical merging to the retrieved subgraphs. Moreover, it provides an ordering among the triplets using the Breadth-First Search (BFS) traversal algorithm. We conducted experiments on the MetaQA benchmark, which was proposed for multi-hop question-answering in the movies domain. Our experiments showed that SG-RAG MOT provided more accurate answers than Chain-of-Thought and Graph Chain-of-Thought. We also found that merging (up to a certain point) highly overlapping subgraphs and defining an order among the triplets helped the LLM to generate more precise answers. Full article

The majority of the challenges faced in power system engineering are presented as constrained optimization functions, which are frequently characterized by their complicated architectures. Metaheuristics are mathematical techniques used to solve complicated optimization problems. One such technique, Artificial Rabbits Optimization (ARO), has been designed to address global optimization challenges. However, ARO has limitations in terms of search functionality, restricting its efficiency in dealing with constrained optimization environments. To improve ARO’s compatibility with a variety of challenging problems, this work proposes implementing the Cauchy mutation operator into the position-updating procedure during the exploration stage. Furthermore, a novel multi-mode control parameter is developed to facilitate a smooth transition between exploration and exploitation phases. The enhancements may boost the performance and serve as an effective optimization tool for tackling complex engineering tasks. The improved version is known as Cauchy Artificial Rabbits Optimization (CARO). The proposed CARO’s performance is evaluated using eleven power system challenges as part of the CEC2020 competition’s test set of real-world constrained problems. The experimental results demonstrate the practical applicability of the proposed CARO in engineering applications and provide areas for future investigation. Full article

Due to the limited endurance of UAVs, especially in scenarios involving large areas and dense target nodes, it is challenging for multiple UAVs to complete diverse tasks while ensuring timely execution. Toward this, we propose a cross-platform system consisting of an aircraft carrier (AC) and multiple UAVs, which makes unified task planning for included heterogeneous platforms to maximize the efficiency of the entire combat system. The carrier-based UAV swarm mission planning problem is formulated to minimize completion time and resource utilization, taking into account large-scale targets, multi-type tasks, and multi-obstacle environments. Since the problem is complex, we design a decoupled framework to simplify the solution by decomposing it into two levels: upper-level AC path planning and bottom-level multi-UAV cooperative mission planning. At the upper level, a drop point determination method and a discrete genetic algorithm incorporating improved A* (DGAIIA) are proposed to plan the AC’s path in the presence of no-fly zones and radar threats. At the bottom level, an improved differential evolution algorithm with a market mechanism (IDEMM) is proposed to minimize task completion time and maximize UAV utilization. Specifically, a dual-switching search strategy and a neighborhood-first buying-and-selling mechanism are developed to improve the search efficiency of the IDEMM. Simulation results validate the effectiveness of both the DGAIIA and IDEMM. An animation of the simulation results is available at simulation section. Full article

Background: Upper-limb impairment is a major cause of post-stroke disability, limiting participation in meaningful activities. Robotic rehabilitation may address this by delivering high-dosage, task-oriented training while reducing clinician workload. However, limited clinical translation of robotic interventions may be partly due to poor reporting in the literature. This systematic review evaluated the intervention-reporting quality (completeness and consistency) of randomized controlled trials (RCTs) comparing robotic and conventional upper-limb stroke rehabilitation. Methods: Four databases were searched for RCTs investigating robotic upper-limb interventions compared with dose-matched conventional interventions for people with stroke. Intervention reporting was assessed using the TIDieR-Rehab checklist. Trained reviewers independently extracted and evaluated data, resolving discrepancies through consensus. Completeness and consistency were analyzed descriptively. Results: Among 25 RCTs, the overall reporting completeness was low (43%). Robotic interventions were better described (50%) than conventional interventions (36%). While timing and total dose were commonly reported, critical details on provider expertise, active dose, progressive challenge, personalization, and harms were often omitted. Reporting consistency was moderate (68%), with key information dispersed across article sections. Conclusions: Inadequate reporting limits the transparency, replication, and implementation of robotic upper-limb interventions. Adopting structured reporting frameworks like TIDieR-Rehab is essential for advancing the field. Full article

Workflow scheduling in cloud computing is attracting increasing attention. Cloud computing can assign tasks to available virtual machine resources in cloud data centers according to scheduling strategies, providing a powerful computing platform for the execution of workflow tasks. However, developing effective workflow scheduling algorithms to find optimal or near-optimal task-to-VM allocation solutions that meet users’ specific QoS requirements still remains an open area of research. In this paper, we propose a hybrid QoS-aware workflow scheduling algorithm named HLWOA to address the problem of simultaneously minimizing the completion time and execution cost of workflow scheduling in cloud computing. First, the workflow scheduling problem in cloud computing is modeled as a multi-objective optimization problem. Then, based on the heterogeneous earliest finish time (HEFT) heuristic optimization algorithm, tasks are reverse topologically sorted and assigned to virtual machines with the earliest finish time to construct an initial workflow task scheduling sequence. Furthermore, an improved Whale Optimization Algorithm (WOA) based on Lévy flight is proposed. The output solution of HEFT is used as one of the initial population solutions in WOA to accelerate the convergence speed of the algorithm. Subsequently, a Lévy flight search strategy is introduced in the iterative optimization phase to avoid the algorithm falling into local optimal solutions. The proposed HLWOA is evaluated on the WorkflowSim platform using real-world scientific workflows (Cybershake and Montage) with different task scales (100 and 1000). Experimental results demonstrate that HLWOA outperforms HEFT, HEPGA, and standard WOA in both makespan and cost, with normalized fitness values consistently ranking first. Full article

Introduction: Hip abductor strength is essential for pelvic stability, lower limb alignment, and injury prevention. Weaknesses of the gluteus medius and minimus contribute to various musculoskeletal conditions. Lateral band walks and monster walks are elastic resistance exercises commonly used to target the hip abductors and external rotators in functional, weight-bearing tasks. Therefore, the aim was to summarize the current evidence on the biomechanics, muscle activation, and clinical applications of lateral and monster band walks. Methods: This narrative review was conducted following the SANRA guideline. A comprehensive literature search was performed across PubMed, Scopus, Web of Science, and SPORTDiscus up to April 2025. Studies on the biomechanics, electromyography, and clinical applications of lateral band walks and monster walks were included, alongside relevant evidence on hip abductor strengthening. Results: A total of 13 studies were included in the review, of which 4 specifically investigated lateral band walk and/or monster walk exercises. Lateral and monster walks elicit moderate to high activation of the gluteus medius and maximus, especially when performed with the band at the ankles or forefeet and in a semi-squat posture. This technique minimizes compensation from the tensor fasciae latae and promotes selective gluteal recruitment. Proper execution requires control of the trunk and pelvis, optimal squat depth, and consistent band tension. Anatomical factors (e.g., femoral torsion), sex differences, and postural variations may influence movement quality and necessitate tailored instruction. Full article

This paper proposes a novel optimization framework for reconfigurable intelligent surface (RIS)-aided movable antenna (MA) systems, tackling the joint optimization problem of beamforming and antenna positions. Unlike traditional approaches, we reformulate the antenna positioning task as a sequential quadratic programming (SQP) problem, enabling efficient handling of nonlinear spatial constraints through iteratively solved quadratic subproblems. An alternating optimization scheme is adopted to decouple the overall problem into two subproblems: (1) optimal beamforming using maximum ratio transmission (MRT) and fixed-point iteration, and (2) precise antenna location optimization via SQP. Simulation results demonstrate that the proposed method significantly enhances spectral efficiency by fully exploiting the synergistic benefits of RIS and MA technologies. The proposed method could achieve about a 25% performance improvement compared to the fixed-position scheme. Current approaches predominantly rely on gradient search methods, which fail to fully exploit the potential of positional DoFs. In contrast, our proposed method is more effective. Full article

Visual object tracking is one of the core techniques in human-centered artificial intelligence, which is very useful for human–machine interaction. State-of-the-art tracking methods have shown their robustness and accuracy on many challenges. However, a large amount of videos with precisely dense annotations are required for fully supervised training of their models. Considering that annotating videos frame-by-frame is a labor- and time-consuming workload, reducing the reliance on manual annotations during the tracking models’ training is an important problem to be resolved. To make a trade-off between the annotating costs and the tracking performance, we propose a weakly supervised tracking method based on co-saliency learning, which can be flexibly integrated into various tracking frameworks to reduce annotation costs and further enhance the target representation in current search images. Since our method enables the model to explore valuable visual information from unlabeled frames, and calculate co-salient attention maps based on multiple frames, our weakly supervised methods can obtain competitive performance compared to fully supervised baseline trackers, using only 3.33% of manual annotations. We integrate our method into two CNN-based trackers and a Transformer-based tracker; extensive experiments on four general tracking benchmarks demonstrate the effectiveness of our method. Furthermore, we also demonstrate the advantages of our method on egocentric tracking task; our weakly supervised method obtains 0.538 success on TREK-150, which is superior to prior state-of-the-art fully supervised tracker by 7.7%. Full article

(1) Medical training is changing, especially for surgeons. Virtual reality simulation is an excellent way to train surgeons safely. Studies show that surgeons who train with simulation have demonstrated improved technical skills in fundamental surgical procedures. The purpose of this study is to determine the overall impact of virtual reality training on laparoscopic cholecystectomy performance and to explore whether specific training protocols or the addition of feedback confer any advantages for future surgeons. (2) MEDLINE (PubMed), Embase (Ovid SP), Web of Science, Google Scholar, and Scopus were searched for the literature related to virtual reality training, immersive simulation, laparoscopic surgical skills training, and medical education. Study quality was assessed using the Cochrane Risk of Bias Tool and NIH Quality Assessment Tool. (3) A total of 55 full-text articles were reviewed. Meta-analysis showed that virtual reality training is an effective method for learning cholecystectomy surgical skills. (4) Conclusions: Performance, measured by objective structured assessments and time to task completion, is improved with virtual reality training compared with no additional training. Positive effects of simulation training were evident in global rating scores and operative time. Continuous feedback on movement parameters during laparoscopic cholecystectomy skills training impacts skills acquisition and long-term retention. Full article

Introduction: Burnout syndrome, long described as an “occupational phenomenon”, now affects 15–20% of the general workforce and more than 50% of clinicians, teachers, social-care staff and first responders. Its precise nosological standing remains disputed. We conducted a mechanistic review of electroencephalography (EEG) studies to determine whether burnout is accompanied by reproducible brain-function alterations that justify disease-level classification. Methods: Following PRISMA-adapted guidelines, two independent reviewers searched PubMed/MEDLINE, Scopus, Google Scholar, Cochrane Library and reference lists (January 1980–May 2025) using combinations of “burnout,” “EEG”, “electroencephalography” and “event-related potential.” Only English-language clinical investigations were eligible. Eighteen studies (n = 2194 participants) met the inclusion criteria. Data were synthesised across three domains: resting-state spectra/connectivity, event-related potentials (ERPs) and longitudinal change. Results: Resting EEG consistently showed (i) a 0.4–0.6 Hz slowing of individual-alpha frequency, (ii) 20–35% global alpha-power reduction and (iii) fragmentation of high-alpha (11–13 Hz) fronto-parietal coherence, with stage- and sex-dependent modulation. ERP paradigms revealed a distinctive “alarm-heavy/evaluation-poor” profile; enlarged N2 and ERN components signalled hyper-reactive conflict and error detection, whereas P3b, Pe, reward-P3 and late CNV amplitudes were attenuated by 25–50%, indicating depleted evaluative and preparatory resources. Feedback processing showed intact or heightened FRN but blunted FRP, and affective tasks demonstrated threat-biassed P3a latency shifts alongside dampened VPP/EPN to positive cues. These alterations persisted in longitudinal cohorts yet normalised after recovery, supporting trait-plus-state dynamics. The electrophysiological fingerprint differed from major depression (no frontal-alpha asymmetry, opposite connectivity pattern). Conclusions: Across paradigms, burnout exhibits a coherent neurophysiological signature comparable in magnitude to established psychiatric disorders, refuting its current classification as a non-disease. Objective EEG markers can complement symptom scales for earlier diagnosis, treatment monitoring and public-health surveillance. Recognising burnout as a clinical disorder—and funding prevention and care accordingly—is medically justified and economically imperative. Full article

Unmanned aerial vehicles (UAVs), commonly known as drones, offer unprecedented flexibility for complex missions such as area surveillance, search and rescue, and cooperative inspection. This paper presents a unified evaluation framework for the comparison of centralized and distributed task allocation algorithms specifically tailored to multi-UAV operations. We first contextualize the classical assignment problem (AP) under UAV mission constraints, including the flight time, propulsion energy capacity, and communication range, and evaluate optimal one-to-one solvers including the Hungarian algorithm, the Bertsekas $ϵ$-auction algorithm, and a minimum cost maximum flow formulation. To reflect the dynamic, uncertain environments that UAV fleets encounter, we extend our analysis to distributed multi-UAV task allocation (MUTA) methods. In particular, we examine the consensus-based bundle algorithm (CBBA) and a distributed auction 2-opt refinement strategy, both of which iteratively negotiate task bundles across UAVs to accommodate real-time task arrivals and intermittent connectivity. Finally, we outline how reinforcement learning (RL) can be incorporated to learn adaptive policies that balance energy efficiency and mission success under varying wind conditions and obstacle fields. Through simulations incorporating UAV-specific cost models and communication topologies, we assess each algorithm’s mission completion time, total energy expenditure, communication overhead, and resilience to UAV failures. Our results highlight the trade-off between strict optimality, which is suitable for small fleets in static scenarios, and scalable, robust coordination, necessary for large, dynamic multi-UAV deployments. Full article

Object detection in aerial images can provide great support in traffic planning, national defense reconnaissance, hydrographic surveys, infrastructure construction, and other fields. Objects in aerial images are characterized by small pixel–area ratios, dense arrangements between objects, and arbitrary inclination angles. In response to these characteristics and problems, we improved the feature extraction network Inception-ResNet using the Fast Architecture Search (FAS) module and proposed a one-stage anchor-free rotation object detector. The structure of the object detector is simple and only consists of convolution layers, which reduces the number of model parameters. At the same time, the label sampling strategy in the training process is optimized to resolve the problem of insufficient sampling. Finally, a decoupled object detection head is used to separate the bounding box regression task from the object classification task. The experimental results show that the proposed method achieves mean average precision (mAP) of 82.6%, 79.5%, and 89.1% on the DOTA1.0, DOTA1.5, and HRSC2016 datasets, respectively, and the detection speed reaches 24.4 FPS, which can meet the needs of real-time detection. Full article

This study introduces a novel approach for minimizing assembly errors in wind turbine gearboxes using a hybrid optimization algorithm, Particle Swarm-Bacteria Foraging Optimization (PSBFO). By integrating error-driven task sequencing and real-time error feedback with the PSBFO algorithm, we developed a comprehensive framework tailored to the unique challenges of gearbox assembly. The PSBFO algorithm combines the global search capabilities of PSO with the local refinement of BFO, creating a unified framework that efficiently explores task sequencing, minimizing misalignment and torque misapplication assembly errors. The methodology results in a 38% reduction in total assembly errors, improving both process accuracy and efficiency. Specifically, the PSBFO algorithm reduced errors from an initial value of 50 to a final value of 5 across 20 iterations, with components such as the low-speed shaft and planetary gear system showing the most substantial reductions. The 50 to 5 error reduction represents a significant decrease in assembly errors from an unoptimized (50) to an optimized (5) sequence, achieved through the PSBFO algorithm, by minimizing dimensional deviations, torque mismatches, and alignment errors across 26 critical gearbox components. While the primary focus is on wind turbine gearbox applications, this approach has the potential for broader applicability in error-prone assembly processes in industries such as automotive and aerospace, warranting further validation in future studies. Full article

Automated generative design is increasingly used across engineering disciplines to accelerate innovation and reduce costs. Generative design offers the prospect of simplifying manual design tasks by exploring the efficacy of solutions automatically. However, existing generative design frameworks rely heavily on expensive optimisation procedures and often produce customised solutions, lacking reusable generative rules that transfer across different problems. This work presents a bioinspired generative design algorithm utilising the concept of evolutionary development (EvoDevo). This evolves a set of developmental rules that can be applied to different engineering problems to rapidly develop designs without the need to run full optimisation procedures. In this approach, an initial design is decomposed into simple entities called cells, which independently control their local growth over a development cycle. In biology, the growth of cells is governed by a gene regulatory network (GRN), but there is no single widely accepted model for this in artificial systems. The GRN responds to the state of the cell induced by external stimuli in its environment, which, in this application, is the loading regime on a bridge truss structure (but can be generalised to any engineering structure). Two GRN models are investigated: graph neural network (GNN) and graph-based Cartesian genetic programming (CGP) models. Both GRN models are evolved using a novel genetic search algorithm for parameter search, which can be re-used for other design problems. It is revealed that the CGP-based method produces results similar to those obtained using the GNN-based methods while offering more interpretability. In this work, it is shown that this EvoDevo approach is able to produce near-optimal truss structures via growth mechanisms such as moving vertices or changing edge features. The technique can be set up to provide design automation for a range of engineering design tasks. Full article

The Grasshopper Optimization Algorithm (GOA) has attracted significant attention due to its simplicity and effective search capabilities. However, its performance deteriorates when dealing with high-dimensional or complex optimization tasks. To address these limitations, this study proposes an improved variant of GOA, named Outpost Multi-population GOA (OMGOA). OMGOA integrates two novel mechanisms: the Outpost mechanism, which enhances local exploitation by guiding agents towards high-potential regions, and the multi-population enhanced mechanism, which promotes global exploration and maintains population diversity through parallel evolution and controlled information exchange. Comprehensive experiments were conducted to evaluate the effectiveness of OMGOA. Ablation studies were performed to assess the individual contributions of each mechanism, while multi-dimensional testing was used to verify robustness and scalability. Comparative experiments show that OMGOA has better optimization performance compared to other similar algorithms. In addition, OMGOA was successfully applied to a real-world engineering problem—lithology prediction from petrophysical logs—where it achieved competitive classification performance. Full article