Selected work

Representative Papers

An Addendum to NeBula: Toward Extending Team CoSTAR’s Solution to Larger Scale Environments

Apr 18, 2025IEEE Transactions on Field Robotics

Autonomous collaborative exploration in ultra-large-scale, unstructured underground environments remains challenging due to severe communication constraints, navigation uncertainty, and lack of prior maps. Method: This work extends TEAM CoSTAR’s NeBula autonomy system with a full-stack enhancement framework integrating semantic-geometric joint mapping, distributed POMDP-based global planning under communication constraints, adaptive filtering for localization, Gaussian process–based probabilistic traversability modeling, edge-cloud cooperative communication protocols, and aerial-ground heterogeneous multi-agent task allocation. Contribution/Results: The framework achieves, for the first time, robust long-range mapping (>5 km²), sub-meter localization accuracy (<0.3 m), and decentralized collaborative decision-making in kilometer-scale underground spaces (e.g., limestone mines). Validated in the DARPA Subterranean Challenge and real-world mine deployments, it improves mission completion rate by 37%, significantly advancing scalability, robustness, and coordination in autonomous underground exploration.

6 citationsRead paper

From User Preferences to Base Score Extraction Functions in Gradual Argumentation

Feb 16, 2026

This study addresses the problem of automatically extracting intrinsic weights from users’ preferences over arguments to construct interpretable gradual argumentation systems. It formally introduces, for the first time, a base weight extraction function that maps qualitative preferences to quantitative base weights for building Quantitative Bipolar Argumentation Frameworks (QBAFs). By incorporating the nonlinear characteristics of human preferences, the approach enables transparent and contestable AI decision-making. Integrating bipolar argumentation frameworks, a base weight extraction algorithm, and gradual semantics, the work provides both theoretical analysis and empirical validation through robotic experiments, demonstrating its effectiveness. Furthermore, it offers practical guidance for selecting appropriate gradual semantics in real-world applications.

1 citationsRead paper

Unsupervised Modular Adaptive Region Growing and RegionMix Classification for Wind Turbine Segmentation

Jan 07, 2026arXiv.org

This work addresses the challenge of scaling pixel-level segmentation of wind turbine blades for automated inspection, which typically relies on extensive manual annotations. To overcome this limitation, the authors propose a fully unsupervised and interpretable modular approach that reformulates the segmentation task as region generation followed by binary classification. The method leverages adaptive region growing—incorporating adaptive thresholding and region merging—and introduces RegionMix, a novel region-mixing augmentation strategy, to significantly enhance model generalization. Evaluated across multiple wind farm datasets, the proposed approach achieves state-of-the-art segmentation accuracy and demonstrates strong cross-site generalization capabilities without requiring any labeled data.

1 citationsRead paper

VisionAssist: An Open-Source Smartphone Assistant for AI-Based Visual Accessibility

Jul 18, 2026

This work addresses the challenges faced by individuals with low vision in independently performing daily tasks due to limited access to visual information. The authors propose and implement an open-source smartphone application that, for the first time, integrates multiple AI-powered visual assistance functionalities—including real-time object localization, image-to-speech description generation, and contextual integration with calendar and contact data—within a unified platform supporting fully voice-driven, hands-free interaction. By synergistically combining computer vision, image recognition, natural language generation, and text-to-speech technologies, the system delivers a comprehensive accessible service. Beyond its practical deployability, the open-source nature of the solution fosters community-driven collaboration, thereby advancing assistive technology innovation and significantly enhancing autonomy for visually impaired users.

0 citationsRead paper

Assessing Physical Frailty and Fall-Risk Indicators with Social Robots: An in situ Evaluation with Older Adults

Jul 16, 2026

This study addresses the limitations of traditional frailty assessments, which rely on coarse-grained clinical indicators, overlook biomechanical signatures of functional decline, and incur high costs. The authors propose an automated frailty evaluation system leveraging a social robot integrated with a behavior tree architecture to coordinate perception, decision-making, and interaction modules. While guiding older adults through standardized clinical tests such as the Short Physical Performance Battery (SPPB) and Timed Up-and-Go (TUG), the system simultaneously captures multimodal mobility metrics—including vision-based skeletal tracking, inertial measurement unit (IMU) data, and gait analysis. This work represents the first integration of social robotics and behavior trees for frailty assessment, enabling efficient and objective quantification of functional status in real-world clinical settings. Experimental results demonstrate strong agreement between robot-derived task durations and gait parameters with clinical gold standards (ICC > 0.9), and substantial concordance in SPPB total scores compared to therapist ratings (κ = 0.67), confirming the system’s reliability and clinical applicability.

0 citationsRead paper

Recent publications

Latest Papers

VisionAssist: An Open-Source Smartphone Assistant for AI-Based Visual Accessibility

Jul 18, 2026

This work addresses the challenges faced by individuals with low vision in independently performing daily tasks due to limited access to visual information. The authors propose and implement an open-source smartphone application that, for the first time, integrates multiple AI-powered visual assistance functionalities—including real-time object localization, image-to-speech description generation, and contextual integration with calendar and contact data—within a unified platform supporting fully voice-driven, hands-free interaction. By synergistically combining computer vision, image recognition, natural language generation, and text-to-speech technologies, the system delivers a comprehensive accessible service. Beyond its practical deployability, the open-source nature of the solution fosters community-driven collaboration, thereby advancing assistive technology innovation and significantly enhancing autonomy for visually impaired users.

0 citationsRead paper

Assessing Physical Frailty and Fall-Risk Indicators with Social Robots: An in situ Evaluation with Older Adults

Jul 16, 2026

This study addresses the limitations of traditional frailty assessments, which rely on coarse-grained clinical indicators, overlook biomechanical signatures of functional decline, and incur high costs. The authors propose an automated frailty evaluation system leveraging a social robot integrated with a behavior tree architecture to coordinate perception, decision-making, and interaction modules. While guiding older adults through standardized clinical tests such as the Short Physical Performance Battery (SPPB) and Timed Up-and-Go (TUG), the system simultaneously captures multimodal mobility metrics—including vision-based skeletal tracking, inertial measurement unit (IMU) data, and gait analysis. This work represents the first integration of social robotics and behavior trees for frailty assessment, enabling efficient and objective quantification of functional status in real-world clinical settings. Experimental results demonstrate strong agreement between robot-derived task durations and gait parameters with clinical gold standards (ICC > 0.9), and substantial concordance in SPPB total scores compared to therapist ratings (κ = 0.67), confirming the system’s reliability and clinical applicability.

0 citationsRead paper

Vision-Based Dribbling for Humanoid Soccer via Privileged Representation Learning

Jul 14, 2026

This work addresses the challenge of enabling humanoid robots to perform dribbling maneuvers in dynamic adversarial environments using only onboard vision. The authors propose an end-to-end reinforcement learning framework that integrates perception and control, embedding a temporal depth image encoder and a task-specific projection layer directly into the policy network. This architecture learns dribbling policies directly from raw visual inputs without relying on explicit state estimation or privileged information. Evaluated on the Booster T1 humanoid robot in simulation, the method achieves a 100% success rate in goal-oriented dribbling tasks, 96% in scenarios with static obstacles, and 46% against actively defending opponents, demonstrating substantially improved robustness and effectiveness under occlusions, rapid ball motion, and complex physical interactions.

0 citationsRead paper

MaxSAT-Based Feedback for Guiding Vision-Language Models in Sudoku

Jul 14, 2026

This work addresses the tendency of vision-language models to produce invalid solutions in structured visual reasoning tasks such as Sudoku, due to insufficient enforcement of logical constraints. The authors propose a neuro-symbolic approach that integrates a MaxSAT solver into the vision-language reasoning pipeline for the first time. Specifically, candidate answers generated by the model are encoded as soft clauses and combined with the hard constraints inherent to Sudoku, forming a MaxSAT problem. Structured feedback from the solver is then used to iteratively refine the model’s outputs. Evaluated across multiple open- and closed-source vision-language models, this method substantially improves logical consistency and full-solution accuracy, with particularly strong performance under full-board optimization settings.

0 citationsRead paper

Responsible Personalisation: The Double-Edged Sword of Personalisation in Human-Robot Interaction

Jul 07, 2026

This study addresses the ethical risks—such as diminished user autonomy, biased user modeling, manipulation, dehumanization, and privacy violations—that arise from personalized human-AI interaction, noting the absence of a systematic, context-sensitive analytical framework in current research. To bridge this gap, the paper proposes an embodied risk analysis framework that integrates the interaction lifecycle with contextual characteristics (e.g., short-term vs. long-term, open-domain vs. closed-domain). By uniquely aligning personalization stages with interaction dynamics, the framework leverages integrative literature synthesis, contextual categorization modeling, and ethical risk mapping to establish a structured evaluation system. It not only elucidates the mechanisms underlying risk evolution across contexts but also yields actionable design guidelines and delineates promising directions for future research.

0 citationsRead paper