Xian Yeow Lee

Jiabin Lin, Xian Yeow Lee, Talukder Jubery et al.

Many physical systems have underlying safety considerations that require that the strategy deployed ensures the satisfaction of a set of constraints. Further, often we have only partial information on the state of the system. We study the problem of safe real-time decision making under uncertainty. In this paper, we formulate a conservative stochastic contextual bandit formulation for real-time decision making when an adversary chooses a distribution on the set of possible contexts and the learner is subject to certain safety/performance constraints. The learner observes only the context distribution and the exact context is unknown, and the goal is to develop an algorithm that selects a sequence of optimal actions to maximize the cumulative reward without violating the safety constraints at any time step. By leveraging the UCB algorithm for this setting, we propose a conservative linear UCB algorithm for stochastic bandits with context distribution. We prove an upper bound on the regret of the algorithm and show that it can be decomposed into three terms: (i) an upper bound for the regret of the standard linear UCB algorithm, (ii) a constant term (independent of time horizon) that accounts for the loss of being conservative in order to satisfy the safety constraint, and (ii) a constant term (independent of time horizon) that accounts for the loss for the contexts being unknown and only the distribution being known. To validate the performance of our approach we perform extensive simulations on synthetic data and on real-world maize data collected through the Genomes to Fields (G2F) initiative.

Zhanhong Jiang, Aditya Balu, Xian Yeow Lee et al.

Regret has been widely adopted as the metric of choice for evaluating the performance of online optimization algorithms for distributed, multi-agent systems. However, data/model variations associated with agents can significantly impact decisions and requires consensus among agents. Moreover, most existing works have focused on developing approaches for (either strongly or non-strongly) convex losses, and very few results have been obtained regarding regret bounds in distributed online optimization for general non-convex losses. To address these two issues, we propose a novel composite regret with a new network regret-based metric to evaluate distributed online optimization algorithms. We concretely define static and dynamic forms of the composite regret. By leveraging the dynamic form of our composite regret, we develop a consensus-based online normalized gradient (CONGD) approach for pseudo-convex losses, and it provably shows a sublinear behavior relating to a regularity term for the path variation of the optimizer. For general non-convex losses, we first shed light on the regret for the setting of distributed online non-convex learning based on recent advances such that no deterministic algorithm can achieve the sublinear regret. We then develop the distributed online non-convex optimization with composite regret (DINOCO) without access to the gradients, depending on an offline optimization oracle. DINOCO is shown to achieve sublinear regret; to our knowledge, this is the first regret bound for general distributed online non-convex learning.

Xian Yeow Lee, Haiyan Wang, Daisuke Katsumata et al.

This paper proposes a multi-agent reinforcement learning (MARL) approach to learn dynamic dispatching strategies, which is crucial for optimizing throughput in material handling systems across diverse industries. To benchmark our method, we developed a material handling environment that reflects the complexities of an actual system, such as various activities at different locations, physical constraints, and inherent uncertainties. To enhance exploration during learning, we propose a method to integrate domain knowledge in the form of existing dynamic dispatching heuristics. Our experimental results show that our method can outperform heuristics by up to 7.4 percent in terms of median throughput. Additionally, we analyze the effect of different architectures on MARL performance when training multiple agents with different functions. We also demonstrate that the MARL agents performance can be further improved by using the first iteration of MARL agents as heuristics to train a second iteration of MARL agents. This work demonstrates the potential of applying MARL to learn effective dynamic dispatching strategies that may be deployed in real-world systems to improve business outcomes.

Aman Kumar, Ekant Muljibhai Amin, Xian Yeow Lee et al.

Large Language Models (LLMs) have shown remarkable success in supporting a wide range of knowledge-intensive tasks. In specialized domains, there is growing interest in leveraging LLMs to assist subject matter experts with domain-specific challenges. However, deploying LLMs as SaaS solutions raises data privacy concerns, while many open-source models demand significant computational resources for effective domain adaptation and deployment. A promising alternative is to develop smaller, domain-specialized LLMs, though this approach is often constrained by the lack of high-quality domain-specific training data. In this work, we address these limitations by presenting a cost-efficient and scalable training pipeline that combines guided synthetic data generation from a small seed corpus with bottom-up domain data curation. Our pipeline integrates Domain-Adaptive Pretraining (DAPT), Domain-specific Supervised Fine-tuning (DSFT), and Direct Preference Optimization (DPO) to train effective small-scale models for specialized use cases. We demonstrate this approach through DiagnosticSLM, a 3B-parameter domain-specific model tailored for fault diagnosis, root cause analysis, and repair recommendation in industrial settings. To evaluate model performance, we introduce four domain-specific benchmarks: multiple-choice questions (DiagnosticMCQ), question answering (DiagnosticQA), sentence completion (DiagnosticComp), and summarization (DiagnosticSum). DiagnosticSLM achieves up to 25% accuracy improvement over open-source models of comparable or larger size (2B-9B) on the MCQ task, while also outperforming or matching them in other tasks, demonstrating effective domain-specific reasoning and generalization capabilities.

Ting-Han Fan, Xian Yeow Lee, Yubo Wang

We introduce PowerGym, an open-source reinforcement learning environment for Volt-Var control in power distribution systems. Following OpenAI Gym APIs, PowerGym targets minimizing power loss and voltage violations under physical networked constraints. PowerGym provides four distribution systems (13Bus, 34Bus, 123Bus, and 8500Node) based on IEEE benchmark systems and design variants for various control difficulties. To foster generalization, PowerGym offers a detailed customization guide for users working with their distribution systems. As a demonstration, we examine state-of-the-art reinforcement learning algorithms in PowerGym and validate the environment by studying controller behaviors. The repository is available at \url{https://github.com/siemens/powergym}.

Xian Yeow Lee, Lasitha Vidyaratne, Gregory Sin et al.

Autonomous inspection in hazardous environments requires AI agents that can interpret high-level goals and execute precise control. A key capability for such agents is spatial grounding, for example when a drone must center a detected object in its camera view to enable reliable inspection. While large language models provide a natural interface for specifying goals, using them directly for visual control achieves only 58\% success in this task. We envision that equipping agents with a world model as a tool would allow them to roll out candidate actions and perform better in spatially grounded settings, but conventional world models are data and compute intensive. To address this, we propose a task-specific latent dynamics model that learns state-specific action-induced shifts in a shared latent space using only goal-state supervision. The model leverages global action embeddings and complementary training losses to stabilize learning. In experiments, our approach achieves 71\% success and generalizes to unseen images and instructions, highlighting the potential of compact, domain-specific latent dynamics models for spatial alignment in autonomous inspection.

Ethan Herron, Xian Yeow Lee, Gregory Sin et al.

Autonomous inspection systems are essential for ensuring the performance and longevity of industrial assets. Recently, agentic frameworks have demonstrated significant potential for automating inspection workflows but have been limited to digital tasks. Their application to physical assets in real-world environments, however, remains underexplored. In this work, our contributions are two-fold: first, we propose a hierarchical agentic framework for autonomous drone control, and second, a reasoning methodology for individual function executions which we refer to as ReActEval. Our framework focuses on visual inspection tasks in indoor industrial settings, such as interpreting industrial readouts or inspecting equipment. It employs a multi-agent system comprising a head agent and multiple worker agents, each controlling a single drone. The head agent performs high-level planning and evaluates outcomes, while worker agents implement ReActEval to reason over and execute low-level actions. Operating entirely in natural language, ReActEval follows a plan, reason, act, evaluate cycle, enabling drones to handle tasks ranging from simple navigation (e.g., flying forward 10 meters and land) to complex high-level tasks (e.g., locating and reading a pressure gauge). The evaluation phase serves as a feedback and/or replanning stage, ensuring actions align with user objectives while preventing undesirable outcomes. We evaluate the framework in a simulated environment with two worker agents, assessing performance qualitatively and quantitatively based on task completion across varying complexity levels and workflow efficiency. By leveraging natural language processing for agent communication, our approach offers a novel, flexible, and user-accessible alternative to traditional drone-based solutions, enabling autonomous problem-solving for industrial inspection without extensive user intervention.

Xian Yeow Lee, Lasitha Vidyaratne, Ahmed Farahat et al.

Large Language Model (LLM)-based systems present new opportunities for autonomous health monitoring in sensor-rich industrial environments. This study explores the potential of LLMs to detect and classify faults directly from sensor data, while producing inherently explainable outputs through natural language reasoning. We systematically evaluate how LLM-system architecture (single-LLM vs. multi-LLM), input representations (raw vs. descriptive statistics), and context window size affect diagnostic performance. Our findings show that LLM systems perform most effectively when provided with summarized statistical inputs, and that systems with multiple LLMs using specialized prompts offer improved sensitivity for fault classification compared to single-LLM systems. While LLMs can produce detailed and human-readable justifications for their decisions, we observe limitations in their ability to adapt over time in continual learning settings, often struggling to calibrate predictions during repeated fault cycles. These insights point to both the promise and the current boundaries of LLM-based systems as transparent, adaptive diagnostic tools in complex environments.

Xian Yeow Lee, Haiyan Wang, Daisuke Katsumata et al.

Dynamic dispatching rules that allocate resources to tasks in real-time play a critical role in ensuring efficient operations of many automated material handling systems across industries. Traditionally, the dispatching rules deployed are typically the result of manually crafted heuristics based on domain experts' knowledge. Generating these rules is time-consuming and often sub-optimal. As enterprises increasingly accumulate vast amounts of operational data, there is significant potential to leverage this big data to enhance the performance of automated systems. One promising approach is to use Decision Transformers, which can be trained on existing enterprise data to learn better dynamic dispatching rules for improving system throughput. In this work, we study the application of Decision Transformers as dynamic dispatching policies within an actual multi-agent material handling system and identify scenarios where enterprises can effectively leverage Decision Transformers on existing big data to gain business value. Our empirical results demonstrate that Decision Transformers can improve the material handling system's throughput by a considerable amount when the heuristic originally used in the enterprise data exhibits moderate performance and involves no randomness. When the original heuristic has strong performance, Decision Transformers can still improve the throughput but with a smaller improvement margin. However, when the original heuristics contain an element of randomness or when the performance of the dataset is below a certain threshold, Decision Transformers fail to outperform the original heuristic. These results highlight both the potential and limitations of Decision Transformers as dispatching policies for automated industrial material handling systems.

Xian Yeow Lee, Aman Kumar, Lasitha Vidyaratne et al.

This paper focuses on solving a fault detection problem using multivariate time series of vibration signals collected from planetary gearboxes in a test rig. Various traditional machine learning and deep learning methods have been proposed for multivariate time-series classification, including distance-based, functional data-oriented, feature-driven, and convolution kernel-based methods. Recent studies have shown using convolution kernel-based methods like ROCKET, and 1D convolutional neural networks with ResNet and FCN, have robust performance for multivariate time-series data classification. We propose an ensemble of three convolution kernel-based methods and show its efficacy on this fault detection problem by outperforming other approaches and achieving an accuracy of more than 98.8\%.

Zhanhong Jiang, Xian Yeow Lee, Sin Yong Tan et al.

We propose a novel policy gradient method for multi-agent reinforcement learning, which leverages two different variance-reduction techniques and does not require large batches over iterations. Specifically, we propose a momentum-based decentralized policy gradient tracking (MDPGT) where a new momentum-based variance reduction technique is used to approximate the local policy gradient surrogate with importance sampling, and an intermediate parameter is adopted to track two consecutive policy gradient surrogates. Moreover, MDPGT provably achieves the best available sample complexity of $\mathcal{O}(N^{-1}ε^{-3})$ for converging to an $ε$-stationary point of the global average of $N$ local performance functions (possibly nonconcave). This outperforms the state-of-the-art sample complexity in decentralized model-free reinforcement learning, and when initialized with a single trajectory, the sample complexity matches those obtained by the existing decentralized policy gradient methods. We further validate the theoretical claim for the Gaussian policy function. When the required error tolerance $ε$ is small enough, MDPGT leads to a linear speed up, which has been previously established in decentralized stochastic optimization, but not for reinforcement learning. Lastly, we provide empirical results on a multi-agent reinforcement learning benchmark environment to support our theoretical findings.

Xian Yeow Lee, Soumik Sarkar, Yubo Wang

Volt-var control (VVC) is the problem of operating power distribution systems within healthy regimes by controlling actuators in power systems. Existing works have mostly adopted the conventional routine of representing the power systems (a graph with tree topology) as vectors to train deep reinforcement learning (RL) policies. We propose a framework that combines RL with graph neural networks and study the benefits and limitations of graph-based policy in the VVC setting. Our results show that graph-based policies converge to the same rewards asymptotically however at a slower rate when compared to vector representation counterpart. We conduct further analysis on the impact of both observations and actions: on the observation end, we examine the robustness of graph-based policy on two typical data acquisition errors in power systems, namely sensor communication failure and measurement misalignment. On the action end, we show that actuators have various impacts on the system, thus using a graph representation induced by power systems topology may not be the optimal choice. In the end, we conduct a case study to demonstrate that the choice of readout function architecture and graph augmentation can further improve training performance and robustness.

Xian Yeow Lee, Yasaman Esfandiari, Kai Liang Tan et al.

Advances in computing resources have resulted in the increasing complexity of cyber-physical systems (CPS). As the complexity of CPS evolved, the focus has shifted from traditional control methods to deep reinforcement learning-based (DRL) methods for control of these systems. This is due to the difficulty of obtaining accurate models of complex CPS for traditional control. However, to securely deploy DRL in production, it is essential to examine the weaknesses of DRL-based controllers (policies) towards malicious attacks from all angles. In this work, we investigate targeted attacks in the action-space domain, also commonly known as actuation attacks in CPS literature, which perturbs the outputs of a controller. We show that a query-based black-box attack model that generates optimal perturbations with respect to an adversarial goal can be formulated as another reinforcement learning problem. Thus, such an adversarial policy can be trained using conventional DRL methods. Experimental results showed that adversarial policies that only observe the nominal policy's output generate stronger attacks than adversarial policies that observe the nominal policy's input and output. Further analysis reveals that nominal policies whose outputs are frequently at the boundaries of the action space are naturally more robust towards adversarial policies. Lastly, we propose the use of adversarial training with transfer learning to induce robust behaviors into the nominal policy, which decreases the rate of successful targeted attacks by 50%.

Kai Liang Tan, Yasaman Esfandiari, Xian Yeow Lee et al.

Adoption of machine learning (ML)-enabled cyber-physical systems (CPS) are becoming prevalent in various sectors of modern society such as transportation, industrial, and power grids. Recent studies in deep reinforcement learning (DRL) have demonstrated its benefits in a large variety of data-driven decisions and control applications. As reliance on ML-enabled systems grows, it is imperative to study the performance of these systems under malicious state and actuator attacks. Traditional control systems employ resilient/fault-tolerant controllers that counter these attacks by correcting the system via error observations. However, in some applications, a resilient controller may not be sufficient to avoid a catastrophic failure. Ideally, a robust approach is more useful in these scenarios where a system is inherently robust (by design) to adversarial attacks. While robust control has a long history of development, robust ML is an emerging research area that has already demonstrated its relevance and urgency. However, the majority of robust ML research has focused on perception tasks and not on decision and control tasks, although the ML (specifically RL) models used for control applications are equally vulnerable to adversarial attacks. In this paper, we show that a well-performing DRL agent that is initially susceptible to action space perturbations (e.g. actuator attacks) can be robustified against similar perturbations through adversarial training.

Xian Yeow Lee, Sambit Ghadai, Kai Liang Tan et al.

Robustness of Deep Reinforcement Learning (DRL) algorithms towards adversarial attacks in real world applications such as those deployed in cyber-physical systems (CPS) are of increasing concern. Numerous studies have investigated the mechanisms of attacks on the RL agent's state space. Nonetheless, attacks on the RL agent's action space (AS) (corresponding to actuators in engineering systems) are equally perverse; such attacks are relatively less studied in the ML literature. In this work, we first frame the problem as an optimization problem of minimizing the cumulative reward of an RL agent with decoupled constraints as the budget of attack. We propose a white-box Myopic Action Space (MAS) attack algorithm that distributes the attacks across the action space dimensions. Next, we reformulate the optimization problem above with the same objective function, but with a temporally coupled constraint on the attack budget to take into account the approximated dynamics of the agent. This leads to the white-box Look-ahead Action Space (LAS) attack algorithm that distributes the attacks across the action and temporal dimensions. Our results shows that using the same amount of resources, the LAS attack deteriorates the agent's performance significantly more than the MAS attack. This reveals the possibility that with limited resource, an adversary can utilize the agent's dynamics to malevolently craft attacks that causes the agent to fail. Additionally, we leverage these attack strategies as a possible tool to gain insights on the potential vulnerabilities of DRL agents.

Xian Yeow Lee, Aaron Havens, Girish Chowdhary et al.

The security of Deep Reinforcement Learning (Deep RL) algorithms deployed in real life applications are of a primary concern. In particular, the robustness of RL agents in cyber-physical systems against adversarial attacks are especially vital since the cost of a malevolent intrusions can be extremely high. Studies have shown Deep Neural Networks (DNN), which forms the core decision-making unit in most modern RL algorithms, are easily subjected to adversarial attacks. Hence, it is imperative that RL agents deployed in real-life applications have the capability to detect and mitigate adversarial attacks in an online fashion. An example of such a framework is the Meta-Learned Advantage Hierarchy (MLAH) agent that utilizes a meta-learning framework to learn policies robustly online. Since the mechanism of this framework are still not fully explored, we conducted multiple experiments to better understand the framework's capabilities and limitations. Our results shows that the MLAH agent exhibits interesting coping behaviors when subjected to different adversarial attacks to maintain a nominal reward. Additionally, the framework exhibits a hierarchical coping capability, based on the adaptability of the Master policy and sub-policies themselves. From empirical results, we also observed that as the interval of adversarial attacks increase, the MLAH agent can maintain a higher distribution of rewards, though at the cost of higher instabilities.

Xian Yeow Lee, Aditya Balu, Daniel Stoecklein et al.

Microfluidic devices are utilized to control and direct flow behavior in a wide variety of applications, particularly in medical diagnostics. A particularly popular form of microfluidics -- called inertial microfluidic flow sculpting -- involves placing a sequence of pillars to controllably deform an initial flow field into a desired one. Inertial flow sculpting can be formally defined as an inverse problem, where one identifies a sequence of pillars (chosen, with replacement, from a finite set of pillars, each of which produce a specific transformation) whose composite transformation results in a user-defined desired transformation. Endemic to most such problems in engineering, inverse problems are usually quite computationally intractable, with most traditional approaches based on search and optimization strategies. In this paper, we pose this inverse problem as a Reinforcement Learning (RL) problem. We train a DoubleDQN agent to learn from this environment. The results suggest that learning is possible using a DoubleDQN model with the success frequency reaching 90% in 200,000 episodes and the rewards converging. While most of the results are obtained by fixing a particular target flow shape to simplify the learning problem, we later demonstrate how to transfer the learning of an agent based on one target shape to another, i.e. from one design to another and thus be useful for a generic design of a flow shape.