Selected work

Representative Papers

Advancements in Synthetic Data Extraction for Industrial Injection Molding

Nov 11, 2025Portuguese Conference on Artificial Intelligence

In injection molding, acquiring high-fidelity real-world data is time-consuming and costly, severely limiting the generalizability of machine learning models. To address this, we propose an LSTM-based modeling framework that synergistically integrates synthetic and real data. A high-fidelity simulation model of the production process generates physically consistent synthetic data, while a tunable data injection strategy enhances dataset diversity without compromising physical plausibility. This approach alleviates reliance on large-scale labeled real data, significantly improving model robustness and prediction accuracy under complex operating conditions. Experimental results demonstrate that judicious incorporation of synthetic data boosts model performance by 12.7%, while concurrently reducing annotation effort, equipment wear, and material waste. Our work establishes a reusable and scalable data augmentation paradigm for data-scarce industrial applications.

3 citationsRead paper

Very Exciting: Zero-Shot Model Predictive Control of Buildings via Excitation-Based Generalized Transfer Learning Models

Sep 11, 2026

本文通过使用基于激励的泛化预训练模型解决建筑中数据驱动的模型预测控制(MPC)需大量数据和建模努力的问题，实现零样本迁移学习以提高控制性能。

0 citationsRead paper

The Role of Vehicles in Digital Forensic Investigations: A Structured Synthesis of Digital Vehicle Forensic Characteristics

Jun 29, 2026

This study addresses the absence of a systematic framework in digital vehicle forensics (DVF), where evidence is fragmented across in-vehicle systems, mobile devices, manufacturer backends, and third-party services. Through a structured review of academic literature, standards, and real-world cases, this work identifies eight core characteristics of DVF for the first time, incorporates an adversarial perspective, formalizes the initial forensic triage problem, and proposes a feature-driven prioritization workflow. The resulting reproducible conceptual framework clarifies strategies for selecting and correlating evidence sources, significantly enhancing the efficiency and rigor of forensic investigations in accident reconstruction, criminal inquiries, and cybersecurity incident response—while explicitly accounting for safety, legal, and privacy constraints.

0 citationsRead paper

Measurement-Calibrated Multi-Camera Fusion for Vision-Based Indoor Localization

Jun 11, 2026

Indoor visual localization is hindered by detection noise, occlusions, and limited camera coverage, leading to multi-stage uncertainties that existing fusion methods fail to explicitly model. This work proposes a component-level error quantification and calibration mechanism that explicitly characterizes the uncertainty in homography calibration, human detection, and motion tracking, and leverages these estimates to optimize multi-camera fusion weights. By transforming the fusion process from a black-box into an interpretable framework, the method significantly enhances trajectory stability and motion smoothness. Experimental results demonstrate that, while yielding only marginal gains in absolute localization accuracy over single-camera baselines, the proposed strategy effectively reduces trajectory variance and substantially improves the continuity and robustness of motion estimation.

0 citationsRead paper

Real-world and simulated thermal data from 960 residential multi-zone buildings in Central Europe

Jun 01, 2026

Existing building thermal datasets lack sufficient diversity in building types and operational conditions, limiting advances in thermal dynamic modeling, energy efficiency control, and fault diagnosis. This work addresses this gap by constructing and releasing the ThermBuild dataset, which uniquely integrates 15 months of high-resolution (15-minute) measured data from two real residential buildings with three years of TRNSYS-simulated data from 958 residential units. The dataset encompasses diverse heat pump systems, building characteristics, and climatic conditions, and includes multidimensional variables such as heat pump operation, indoor environmental parameters, and weather data. ThermBuild enables research in cross-domain transfer learning, simulation-to-reality generalization, and model benchmarking, significantly enhancing the applicability, robustness, and reproducibility of data-driven approaches in building energy systems.

0 citationsRead paper

Recent publications

Latest Papers

Very Exciting: Zero-Shot Model Predictive Control of Buildings via Excitation-Based Generalized Transfer Learning Models

Sep 11, 2026

本文通过使用基于激励的泛化预训练模型解决建筑中数据驱动的模型预测控制(MPC)需大量数据和建模努力的问题，实现零样本迁移学习以提高控制性能。

0 citationsRead paper

The Role of Vehicles in Digital Forensic Investigations: A Structured Synthesis of Digital Vehicle Forensic Characteristics

Jun 29, 2026

This study addresses the absence of a systematic framework in digital vehicle forensics (DVF), where evidence is fragmented across in-vehicle systems, mobile devices, manufacturer backends, and third-party services. Through a structured review of academic literature, standards, and real-world cases, this work identifies eight core characteristics of DVF for the first time, incorporates an adversarial perspective, formalizes the initial forensic triage problem, and proposes a feature-driven prioritization workflow. The resulting reproducible conceptual framework clarifies strategies for selecting and correlating evidence sources, significantly enhancing the efficiency and rigor of forensic investigations in accident reconstruction, criminal inquiries, and cybersecurity incident response—while explicitly accounting for safety, legal, and privacy constraints.

0 citationsRead paper

Measurement-Calibrated Multi-Camera Fusion for Vision-Based Indoor Localization

Jun 11, 2026

Indoor visual localization is hindered by detection noise, occlusions, and limited camera coverage, leading to multi-stage uncertainties that existing fusion methods fail to explicitly model. This work proposes a component-level error quantification and calibration mechanism that explicitly characterizes the uncertainty in homography calibration, human detection, and motion tracking, and leverages these estimates to optimize multi-camera fusion weights. By transforming the fusion process from a black-box into an interpretable framework, the method significantly enhances trajectory stability and motion smoothness. Experimental results demonstrate that, while yielding only marginal gains in absolute localization accuracy over single-camera baselines, the proposed strategy effectively reduces trajectory variance and substantially improves the continuity and robustness of motion estimation.

0 citationsRead paper

Real-world and simulated thermal data from 960 residential multi-zone buildings in Central Europe

Jun 01, 2026

Existing building thermal datasets lack sufficient diversity in building types and operational conditions, limiting advances in thermal dynamic modeling, energy efficiency control, and fault diagnosis. This work addresses this gap by constructing and releasing the ThermBuild dataset, which uniquely integrates 15 months of high-resolution (15-minute) measured data from two real residential buildings with three years of TRNSYS-simulated data from 958 residential units. The dataset encompasses diverse heat pump systems, building characteristics, and climatic conditions, and includes multidimensional variables such as heat pump operation, indoor environmental parameters, and weather data. ThermBuild enables research in cross-domain transfer learning, simulation-to-reality generalization, and model benchmarking, significantly enhancing the applicability, robustness, and reproducibility of data-driven approaches in building energy systems.

0 citationsRead paper

Bridging the Sim-to-Real Gap in Reinforcement Learning-Based Industrial Dispatching through Execution Semantics

May 27, 2026

This work addresses the challenges in industrial scheduling where asynchronous event streams often lead to inconsistent decision states, ambiguous action validity, and difficulties in attributing execution errors in reinforcement learning policies. To resolve these issues, the paper proposes a policy-decoupled execution and measurement layer that bridges the policy and the execution environment. By constructing valid decision snapshots, defining standardized execution contracts, and recording multidimensional execution deviations, the approach structurally formalizes execution semantics for the first time. This enables observable and attributable deployment discrepancies between simulation and reality, transforming ambiguous execution failures into type-labeled supervisory signals. Experimental results demonstrate that the framework consistently enhances diagnostic capability across varying observation delays, significantly reducing avoidable errors under low-latency conditions and providing structured supervisory data for policy evaluation and optimization.

0 citationsRead paper