Artificial Intelligence and Robotics Commons^™

Metarag: Identifying Website Owner Using Meta-Path-Guided Dynamic Graph Retrieval-Augmented Generation, Cheng Tu, Yunshan Ma, Bingyang Guo, Qianyu Li, Yang Li, Min Zhang, Fan Shi, Xiang Wang

Research Collection School Of Computing and Information Systems

Website owner identification aims to link websites to their real-world owners, which is crucial for credibility assessment and information provenance in information retrieval and vital for applications in cybersecurity, Internet governance, and digital regulation. Existing approaches for website owner identification primarily rely on querying infrastructure registration records or analyzing webpage content. However, these methods often fail due to incomplete or outdated registration records and sparse webpage content. We observe that inter-website relationships, derived from shared infrastructure data such as primary domains, IP blocks, and geolocations, can provide valuable but underutilized ownership cues. To exploit this insight, we propose MetaRAG, a …

Stprompt++: Prompting Vision-Language Models For Weakly Supervised Video Anomaly Detection And Fine-Grained Localization, Peng Wu, Chengyu Pan, Guansong Pang, Xiangteng He, Zhiwei Yang, Peng Wang, Yanning Zhang

Research Collection School Of Computing and Information Systems

Traditional weakly supervised video anomaly detection (WSVAD) tasks typically rely on coarse-grained frame-level labels for training. Although this approach reduces annotation costs, it results in weak semantic understanding and spatial localization capabilities due to the absence of fine-grained annotations, hindering precise pixel-level anomaly detection and localization. Thanks to the success of vision-language models (VLMs), e.g., CLIP, recent approaches leveraging large VLMs focus on exploiting their strong semantic understanding capabilities, but they typically feed only keyframes or short video segments into the models, without supplying sufficient prior contextual information (e.g., contextual frames around anomalies, zoomed-in anomaly regions, and detailed anomaly descriptions), …

Artificial Intelligence And Translation: Exploring Current Applications, Limitations And Future Potential Of Language Models Through Japanese-English Translation, Loklin Elias Nord

Undergraduate Theses, Capstones, and Recitals

This thesis highlights the recent improvements and capabilities of Large Language Models (LLMs), specifically their ability to produce translations between different languages. The continued up-scaling of model sizes has led to breakthroughs in the level of their observed intelligence, allowing them to produce translations that are similar in quality to highly skilled human translators. However, to facilitate the reasoning processes that LLMs now possess, their demand for computational power and the supporting hardware and resources has increased proportionally. Considering the impacts of this technology on the environment, energy resources, and its accessibility, my research explores the possibilities of smaller, highly …

Ai Failures In The Eyes Of The Downstream Developer: A First Look At Concerns, Practices, And Challenges, Haoyu Gao, Mansooreh Zahedi, Wenxin Jiang, Hong Yi Lin, James C. Davis, Christoph Treude

Research Collection School Of Computing and Information Systems

With the advancement of AI models, more software systems are adopting AI as a component to facilitate automation. Pre-trained models (PTMs) have become a cornerstone of AI-based software, allowing for rapid integration and development with lower training cost. However, their adoption also introduces failure modes such as data leakage and biased outputs, that may require careful handling by downstream developers. While previous research has proposed taxonomies of these technical concerns and various mitigation strategies, how downstream developers address these issues during the development of general AI-based software when reusing PTMs remains unexplored. Understanding downstream developers’ perspectives is essential because they …

Logupdater: Automated Detection And Repair Of Specific Defects In Logging Statements, Renyi Zhong, Yichen Li, Jinxi Kuang, Wenwei Gu, Yintong Huo, Michael R. Lyu

Research Collection School Of Computing and Information Systems

Developers write logging statements to monitor software runtime behaviors and system state. However, poorly constructed or misleading log messages can inadvertently obfuscate actual program execution patterns, thereby impeding effective software maintenance. Existing research on analyzing issues within logging statements is limited, primarily focusing on detecting a singular type of defect and relying on manual intervention for fixes rather than automated solutions.To address the limitation, we initiate a systematic study that pinpoints four specific types of defects in logging statements (i.e., statement code inconsistency, static dynamic inconsistency, temporal relation inconsistency, and readability issues) through the analysis of real-world log-centric changes. We …

Generalizing Unsupervised Lidar Motion Estimation Via Patch-Aware Feature Reweighting Under Snowy Conditions, Beibei Zhou, Zhiyuan Zhang, Zhenbo Song, Jianhui Guo, Tongpo Zhang, Shaojing Song, Qinge Wu, Hui Kong

Research Collection School Of Computing and Information Systems

Unsupervised LiDAR motion estimation has attracted significant attention in pattern recognition and robotic vision. However, ensuring robust performance under adverse conditions — particularly structured noise like heavy snowfall — remains a challenging problem. Existing models struggle with domain generalization due to the sensitivity of feature extractors to spatial outliers, limiting their reliability in complex real-world scenarios. In this work, we present an unsupervised LiDAR motion estimation framework via patch-aware feature reweighting to mitigate the domain shift between clear and snowy conditions. Our approach prioritizes reliability-aware reweighting to reduce the influence of snowflake noise and outlier points on pose estimation, while …

How Leaders Build Employee Trust In Artificial Intelligence: Voice Opportunities, Humility, And Trust Transfer, Jack Mcguire, David De Cremer, Devesh Narayanan

Research Collection Lee Kong Chian School Of Business

Artificial intelligence is increasingly central to organizational work, yet employee trust in AI remains fragile. Although prior research has primarily explained trust in AI through technological characteristics such as transparency, reliability, and accuracy, we argue that trust in AI is also shaped by the social context in which employees encounter these systems. Drawing on affect-as-information theory and social information processing theory, we develop and test a model in which leader-provided voice opportunities reduce employees’ negative affect about AI-related work experiences, thereby enhancing perceptions of leader trustworthiness and, in turn, trust in AI. We further propose that this indirect effect depends …

Machine Learning-Based Regression For Magnetic Field Prediction From Odmr Spectral Data, Jesse B. Hernandez

Electronic Theses, Projects, and Dissertations

Optically Detected Magnetic Resonance (ODMR) using nitrogen-vacancy (NV) centers in diamond enables sensitive, room-temperature magnetic field sensing, but real ODMR spectra are often noisy and difficult to analyze with traditional peak-fitting methods. This thesis investigates whether machine learning can reliably predict magnetic field strength directly from ODMR spectra, and compares four model families under a single regression task: a random forest, an artificial neural network (ANN), a one-dimensional convolutional neural network (1D-CNN), and a Transformer.

Training data were generated from an NV-ensemble simulation calibrated to real measurements provided by the Ulsan National Institute of Science and Technology (UNIST), spanning 0 …

Suicide Ideation Detection Using Social Media Data And Ensemble Machine Learning Model, Erol Kina, Jin Ghoo Choi, Abid Ishaq, Rahman Shafique, Monica Gracia Villar, Eduardo Silva Alvarado, Isabel De La Torre Diez, Imran Ashraf

Research outputs 2022 to 2026

Identifying the emotional state of individuals has useful applications, particularly to reduce the risk of suicide. Users’ thoughts on social media platforms can be used to find cues on the emotional state of individuals. Clinical approaches to suicide ideation detection primarily rely on evaluation by psychologists, medical experts, etc., which is time-consuming and requires medical expertise. Machine learning approaches have shown potential in automating suicide detection. In this regard, this study presents a soft voting ensemble model (SVEM) by leveraging random forest, logistic regression, and stochastic gradient descent classifiers using soft voting. In addition, for the robust training of SVEM, …

Generative Ai Adoption And Solvers' Popularity On Supply-Driven Crowdsourcing Platforms: The Dual Role Of Price Signals, Zimeng Zhu, Carol Hsu, Fiona Fui-Hoon Nah, Na Liu

Research Collection School Of Computing and Information Systems

Purpose – We investigate the effect of solvers’ adoption of Generative AI (GenAI) on their popularity in a supply-driven crowdsourcing platform. We also examine the impact of price signals as well as their heterogeneous impact based on the solvers’ membership duration on the platform. Design/methodology/approach – Our analysis focuses on solvers who adopt GenAI for design-related gigs on the supply-driven crowdsourcing platform. By combining propensity score matching (PSM) with multi-period difference-in-differences (DID), we examine how GenAI adoption impacts solvers’ popularity and how price signals affect this main effect. Findings – Our findings reveal that solvers who adopt GenAI tend to …

Understanding The Out-Of-Distribution Generalization Of Chain-Of-Thought Reasoning In Llms, Nuojing Liang, Xiaotong Yuan, Pan Zhou

Research Collection School Of Computing and Information Systems

Chain-of-Thought (CoT) reasoning substantially improves Large Language Models (LLMs), yet its out-of-distribution (OOD) generalization mechanism remains theoretically underexplored. We study CoT as compositional OOD generalization: the target task lacks complete reasoning trajectories or direct input-output pairs, while training provides only in-distribution (ID) atomic step data. We show that CoT generalizes by repeatedly applying a shared single-step predictor to compose reusable ID transitions. Theoretically, the OOD task risk is controlled by the sum of ID subtask risks, scaled by a depth-independent hard-decoding coefficient Bℓ b = O(k) that is independent of the reasoning depth T. A Rademachercomplexity analysis further gives an …

Drag As Evidence: Motion-Grounded Latent Recomposition For Drag-Based Editing, Xinyu Pu, Hongsong Wang, Jie Gui, Pan Zhou

Research Collection School Of Computing and Information Systems

Modern image editors excel at semantic manipulation and visual synthesis, yet remain limited in precise spatial control, motivating the development of drag-based editing. However, existing drag-based methods often struggle to balance drag accuracy with natural, plausible, and intent-aligned generation. We propose MoReDrag, a motion-grounded drag-based editing method. Our key insight is to treat pixel-space warping as coarse motion evidence, and to inject this evidence into the generative sampling trajectory. Specifically, MoRe-Drag performs region-aware latent recomposition over refinement, inpainting, and anchor regions, coupled with stage-adaptive conditioning that progressively shifts from motion-grounded structure formation to semantic refinement. We further support an instruction-free …

Llm-Enabled Applications Require System-Level Threat Monitoring, Yedi Zhang, Haoyu Wang, Xianglin Yang, Jin Song Dong, Jun Sun

Research Collection School Of Computing and Information Systems

LLM-enabled applications are rapidly reshaping the software ecosystem by using large language models as core reasoning components for complex task execution. This paradigm shift, however, introduces fundamentally new reliability challenges and significantly expands the security attack surface, due to the non-deterministic, learning-driven, and difficult-to-verify nature of LLM behavior. In light of these emerging and unavoidable safety challenges, we argue that such risks should be treated as expected operational conditions rather than exceptional events, necessitating a dedicated incident-response perspective. Consequently, the primary barrier to trustworthy deployment is not further improving model capability but establishing system-level threat monitoring mechanisms that can detect …

From Post-Hoc To Pre-Assessment Evaluation: Llms As Collaborative Partners For Simulating Student Performance, Siaw Ling Lo, Ping Fan Ke, Yi Meng Lau

Research Collection School Of Computing and Information Systems

Ensuring the quality of multiple-choice questions (MCQs) remains a complex challenge in higher education, as traditional post-hoc psychometric analyses often identify flaws only after students have been negatively impacted. While Large Language Models (LLMs) offer potential for simulating student performance to validate items before delivery, they are frequently hindered by a “high-accuracy bias” that prevents larger models from realistically mimicking student errors. This paper proposes the Triadic Pedagogical Fit (TPF) framework, which triangulates Model Scale, Task Complexity via Revised Bloom’s Taxonomy, and Instructional Persona to optimize pre-assessment evaluation. Through two studies, we demonstrate that while Mid-scale (12B) models effectively mirror …

Semantic-Structural Decoupling: Disentangling Semantic Attention From Structural Bias In The Attention Manifold, Pengkun Jiao, Bin Zhu, Jingjing Chen, Yu-Gang Jiang

Research Collection School Of Computing and Information Systems

The empirical success of attention mechanism in Multimodal Large Language Models (MLLMs) often obscures its inherent, subtle flaws. Specifically, MLLMs consistently exhibit disproportionate attention toward certain semantically uninformative visual tokens, a phenomenon termed "register" or "Visual Attention Sinks." While existing inference intervention methods attempt to identify these sink tokens and redistribute their attention weights, such approaches typically treat these tokens in isolation and suffer from computational inefficiency. Instead, we reframe this phenomenon as a generalized textual bias exerted over visual features that extends beyond isolated sink tokens. From this perspective, a pervasive structural bias leads to the dilution of the …

Keyframe Selection From Motion Capture Data With Dual-Agent Reinforcement Learning, Kun Hu, Wang, Clinton Mo, Mingyang Ma, Shaohui Mei, Zebin Chen, Zhiyong Wang

Research outputs 2022 to 2026

Animation production workflows centered around motion capture techniques require animators to edit motions based on a set of keyframes. However, most existing keyframe selection methods are optimization-based, which suffer from the issues of flexibility and efficiency. In this paper, a novel deep reinforcement learning method with dual agents are proposed for unsupervised keyframe selection. First, an S-Agent and an R-Agent evaluate the actions of selection and refinement, respectively. A deep spatio-temporal network, namely graph keyframe evaluation network (GKEN), is proposed for the agents. Then, an animation specified reward is devised based on reconstruction, which fulfills three important properties of the …

Spatialimaginer: Towards Adaptive Visual Imagination For Spatial Reasoning, Yian Li, Yang Jiao, Bin Zhu, Tianwen Qian, Shaoxiang Chen, Jingjing Chen, Yu-Gang Jiang

Research Collection School Of Computing and Information Systems

Spatial intelligence, which refers to the ability to reason about geometric and physical structure from visual observations, remains a core challenge for multimodal large language models. Despite promising performance, recent multimodal large language models (MLLMs) often exhibit fragile reasoning traces in spatial intelligence tasks that involve consistent spatial state recognition. We argue that these failures stem from a mismatch between the spatial recognition mechanism and the text-only reasoning behavior of these MLLMs. Effective spatial reasoning requires low-level geometric structure to be faithfully preserved and updated throughout the reasoning process, whereas textual representations tend to abstract away precisely these critical details. …

Aerial Refueling Scheduling Plan Generation Method For Multi-Position Task Management *, Jiachen Hu, Huanyu Li

Journal of System Simulation

Abstract: An aerial refueling scheduling plan generation method based on hierarchical hybrid strategy is proposed to address the high complexity of aerial refueling scheduling for multi-position tasks and the difficulty for traditional manual operations or single-position-oriented solutions to manage tasks effectively. A multi-level strategic framework based on fixed-rule heuristics, rolling horizon optimization, and steady-state cyclic scheduling is developed and integrated into a hierarchical hybrid strategy to balance scheduling efficiency, optimization capability, and emergency response capability. Simulation results demonstrate that the proposed method effectively generates feasible refueling scheduling plans, producing viable plans under both undisturbed rotational operations and disturbed dynamic support …

Flight Test Mission Planning Method Based On Adaptive Large Neighborhood Search, Luofu Wu, Xinyang Wu, Lunhao Ju, Guopeng Song, Rui Wang, Jiang Jiang

Journal of System Simulation

Abstract: Flight test mission planning is critical to aircraft performance evaluation, but traditional manual methods often struggle with extensive solution spaces and complex constraints. This paper constructs an overall flight test mission planning model and introduces a flight test planning algorithm (FTPA) based on adaptive large neighborhood search. FTPA utilizes Kahn topological sorting to generate an initial test-item priority sequence, which is then decoded into an initial solution, and jointly schedules flight test items and test aircraft within the adaptive large neighborhood search framework. A comparison with the CPLEX solver shows that the FTPA algorithm yields solutions with a gap …

Large Language Model-Assisted Control Framework For Mixed Remote Sensing Satellite Constellations, Guohua Wu, Ziyan You, Lei Li, Yi Gu, Shuanglin Li

Journal of System Simulation

Abstract: A LLM-assisted control framework for mixed remote sensing satellite constellations is proposed to address the problem that the control and scheduling of mixed remote sensing satellite constellations cannot process users' raw natural language observation requirements and realize full-process automation from requirement input to scheduling schemes. LLMs are utilized to enhance the construction of domain knowledge graphs, requirement understanding and task characterization, and realize intelligent generation of scheduling schemes and optimal solution of scheduling models. A scheduling scheme generation method based on the "planning and execution" dual-Agent architecture is designed. Combined with the constellation scheduling toolset, the capability boundary of …

On-Orbit Refueling Mission Planning Of Satellites Based On Improved Nsga-Ii, Guozheng Li, Rui Wang, Yi Lu

Journal of System Simulation

Abstract: An improved NSGA-II solution framework based on multi-tree encoding is proposed to address the problems of high modeling complexity under large-scale scenarios and the tendency of traditional algorithms to fall into local optima. Based on multi-tree modeling, the complex mixed-integer programming problem is transformed into a multi-tree configuration optimization problem, thus reducing the constraint handling difficulty and improving the search efficiency of the solution space. Multi-tree crossover and mutation operators as well as the calculation method for objective functions are designed. A large neighborhood search (LNS) strategy is introduced, and the global exploration and local exploitation capabilities of the …

Llm-Driven Simulation Method For Mobilization Decision-Making In Weapons And Equipment Supply Chain, Qichang Guo, Jihai Zhang, Zichu Qin

Journal of System Simulation

Abstract: A LLM-driven multi-agent simulation model for production mobilization in the weapons and equipment supply chain is developed to address the practical challenge of transitioning the supply chain from peacetime to wartime production in response to surging wartime demand. The model consists of a military demand layer, a supply chain execution layer, and a mobilization decision-making layer. An LLM is incorporated as a mobilization decision generation module to represent the dynamic regulation of production conversion and expansion under war-readiness budget and production safety constraints. The simulation results show that compared with the no-mobilization and rule-based mobilization strategies, the LLM-based strategy …

Chemical Plant Control Based On Dynamic Simulation And Reinforcement Learning, Yuxin Li, Haijie Li, Shangbo Han, Kai Gao, Qingyuan Zhang, Yanwen Xiao, Zhengcai Liu

Journal of System Simulation

Abstract: An intelligent control strategy framework combining offline training and on-site deployment for regenerative thermal oxidizer (RTO) operation control based on RL is proposed to solve the problem that fluctuations in exhaust gas concentration and flow, as well as changes in working conditions, make it difficult for RTO to maintain the system operating parameters within the economically optimal range in practical applications. In the offline phase, RL is employed to enable autonomous policy learning and global optimization. In the online phase, an adaptive policy correction mechanism based on real-time data is introduced. RL control strategy is integrated with the operation …

Five-Dimensional Digital Twin System For Closed-Loop Gasoline Blending Optimization, Yaning Fang, Renchu He

Journal of System Simulation

Abstract: An adaptive blending optimization system based on five-dimensional digital twin theory is designed and developed to address the challenges of strict nonlinear constraints and frequent raw material fluctuations. A five-dimensional model architecture is constructed to achieve virtual-real mapping and synergy throughout the blending process. A closed-loop control strategy of "online model correction and dynamic recipe optimization" is proposed. The constrained recursive least squares (RLS) algorithm is adopted to update the octane blending effect model parameters, resolving the model mismatch caused by raw material fluctuations. The linear decreasing weight particle swarm optimization algorithm (LDW-PSO) is combined to determine the optimal …

Research On Digital Shipyard Design-Procurement Coupling Based On Multi-Agent Systems, Kaitai Peng, Dingyi Feng, Wenbo Zhao, Feier Chen

Journal of System Simulation

Abstract: To address the issues of low supply chain collaboration efficiency and inadequate dynamic scheduling capability caused by the disconnection between design and procurement in digital shipbuilding, a method integrating multi-agent simulation with the Q-learning reinforcement learning algorithm is proposed, and a four-tier dynamic supply chain model coupling design and procurement is developed, covering mines, steel mills, shipyards, and shipowners. Based on the depth of collaboration, steel mills are classified into three levels, including full collaboration, partial collaboration, and noncollaboration, with differentiated state spaces and reward functions designed to simulate intelligent pricing and game behaviors under information asymmetry. Simulation results …

Mission-Oriented Combat Effectiveness Evaluation Model For Intelligent Equipment System-Of-Systems, Long Zhang, Wenbo Huang, Xuanming Feng, Zhen Lei, Ying Wang

Journal of System Simulation

Abstract: A hierarchical evolutionary adaptive learning (HEAL) framework is proposed to address the problems in the operational simulation of intelligent equipment systems under complex adversarial environments, including fragmented hierarchical decision-making, unclear mapping between algorithmic indicators and operational effectiveness, and the lack of adaptive evolution capability of simulation models. A multi-layer learning architecture that integrates robust autonomous decision-making for single units, emergent collaboration of system networks, and game-theoretic evolution of system-of-systems strategies is constructed by following the “micro-meso-macro” system-oriented hierarchical modeling paradigm. A cross-layer closed-loop coupling among learned policies, cooperative structures, and operational effectiveness is achieved by combining bottom-up behavioral abstraction …

Multimodal Perception-Based Trajectory Planning Method For Uavs In Forest Environments, Dequan Li, Jiahao Jiang

Journal of System Simulation

Abstract: To address the requirements for low-latency local planning and safe obstacle avoidance in UAV forest-fire patrol under dense tree-trunk occlusion and locally invalid depth observations, this paper proposes a learning-based local trajectory planning method that fuses depth images, local occupancy distance maps, candidate-level collision-risk assessment, and pre-execution safety reranking. The method uses depth images, local occupancy distance maps, motion states, and target direction as inputs and generates 15 quintic-polynomial candidate trajectories in parallel using 3 × 5 angular motion primitives. During training, candidate safety costs, ranking supervision, and near-collision risk labels are constructed. During inference, online LiDAR point clouds …

Manifold Alignment And Dual Prediction-Based Adaptive Dynamic Multi-Objective Evolutionary Algorithm, Xuezhi Yue, Chao Xiong, Yanshen Jiang, Zhaolu Guo, Hu Peng

Journal of System Simulation

Abstract: To address prediction failure, manifold topology distortion, and insufficient adaptability in environmental response and operator configuration in existing prediction-based algorithms for dynamic multi-objective optimization under complex nonlinear environments and small-sample settings, this paper proposes a manifold alignment and dual prediction-based adaptive dynamic multi-objective evolutionary algorithm, termed MADP-DMOEA. When the environment changes, the algorithm decomposes the decision-variable manifold into center and shape components, effectively aligns manifolds at adjacent time instants via bipartite optimal matching, and identifies environmental change severity using geometric displacement. A dual prediction model is developed, in which a GRNN performs center extrapolation under severe changes, while an …

Cooperative Hunting Method For Multiple Usvs Based On Improved Maddpg, Yonghao Zhang, Changyou Xue, Xueman Fan, Zhuo Cheng

Journal of System Simulation

Abstract: A cooperative hunting method based on improved multi-agent deep deterministic policy gradient (MADDPG) is proposed to address the challenges of non-stationary environments, sparse rewards, and partial observations in the cooperative hunting of targets with unknown motion patterns by multiple unmanned surface vessels (USVs). A Markov game model for hunting is built, and a multidimensional guided reward shaping function incorporating collision avoidance, hunting success, and energy consumption penalties is designed to mitigate the sparse reward problem. ResNet residual networks are introduced to construct the Actor and Critic networks and thus enhance the representation capability of deep networks for complex state-action …

Five-Axis Multi-Station Process Planning Based On Multi-Task Graph Reinforcement Learning, Junjie Liu, Jianlei Xu, Litao Ju, Yanhai Hu, Qinghao Zhang, Qiang Liu

Journal of System Simulation

Abstract: A hierarchical intelligent planning architecture integrating multi-task graph and masked proximal policy optimization (MTG-MPPO) is proposed to address the challenges of factorial expansion of the state space and multi-objective interference in five-axis multi-feature machining. An attributed adjacency graph integrating boundary representation and machining direction vectors is constructed. Probabilistic soft labels are output through a multi-task network to quantify process flexibility and thus achieve prior topological pruning of solution space. An MDP model integrating directed acyclic graph masks and dense reward shaping is built. A Monte Carlo (MC) optimization mechanism based on policy distribution is introduced to coordinate cutting, tool-changing, …