arXiv subjects

Junyi Xu

Publications and source records attributed to Junyi Xu.

4 recordsLinked to original sources

arXiv2026

LIVIN: Benchmarking Spatial and Embodied Intelligence in Digital Twins of Lived-In Homes

Realistic household simulation must capture not only diverse environments but also the lived-in object arrangements and spatial constraints that shape robot motion and interaction. Existing resources often trade off scale, real-world correspondence, and interaction readiness, leaving a gap in faithful, interactive replicas of how real homes are actually arranged. To this end, we introduce LIVIN, a benchmark for spatial and embodied intelligence built on digital twins of 30 diverse lived-in homes. These replicas preserve observed room layouts, furniture configurations, and everyday belongings. To construct them, we design a human-in-the-loop workflow comprising instance recognition, architectural reconstruction, and object generation and placement, with intermediate results reviewed and corrected by humans against the source observations at each stage. We evaluate four tasks in LIVIN: 3D detection, 3D reconstruction, navigation, and loco-manipulation. Our evaluations show that current methods remain challenged by the dense object arrangements, occlusions, limited free space, and constrained interaction regions found in realistic lived-in homes. We hope LIVIN will help advance embodied AI in real-world homes, from spatial understanding to robotic interaction, and ultimately bring embodied intelligence into everyday home environments.

cs.CV↗

arXiv2025

Research on Core Loss of Direct-drive 75kW Tidal Current Generator Using Machine Learning and Multi-objective Optimization Algorithms

This paper presents a classification of generator excitation waveforms using principal component analysis (PCA) and machine learning models, including logistic regression, random forest, and gradient boosting decision trees (GBDT). Building upon the traditional Steinmetz equation, a temperature correction term is introduced. Through nonlinear regression and least squares parameter fitting, a novel temperature correction equation is proposed, which significantly reduces the prediction error for core losses under high-temperature conditions. The average relative error is decreased to 16.03%, thereby markedly enhancing the accuracy. Using GBDT and random forest regression models, the independent and combined effects of temperature, excitation waveforms, and magnetic materials on core loss are analyzed. The results indicate that the excitation waveform has the most significant impact, followed by temperature, while the magnetic core material exhibits the least influence. The optimal combination for minimizing core loss is identified, achieving a core loss value of 35,310.9988 under the specified conditions. A data-driven predictive model for core loss is developed, demonstrating excellent performance with an R*R value of 0.9703 through machine learning regression analysis, indicating high prediction accuracy and broad applicability. Furthermore, a multi-objective optimization model considering both core loss and transmission magnetic energy is proposed. Genetic algorithms are employed for optimization, resulting in an optimal design scheme that minimizes core loss and maximizes transmission magnetic energy. Based on this model, the magnetic core compensation structure of a direct-drive 75kW tidal energy generator is optimized in practical applications, yielding satisfactory results.

physics.app-ph↗

arXiv2024

CalTag: Robust calibration of mmWave Radar and LiDAR using backscatter tags

The rise of automation in robotics necessitates the use of high-quality perception systems, often through the use of multiple sensors. A crucial aspect of a successfully deployed multi-sensor system is the calibration with a known object typically named fiducial. In this work, we propose a novel fiducial system for millimeter wave radars, termed as CalTag. CalTag addresses the limitations of traditional corner reflector-based calibration methods in extremely cluttered environments. CalTag leverages millimeter wave backscatter technology to achieve more reliable calibration than corner reflectors, enhancing the overall performance of multi-sensor perception systems. We compare the performance in several real-world environments and show the improvement achieved by using CalTag as the radar fiducial over a corner reflector.

cs.RO↗

arXiv2022

Machine learning for percolation utilizing auxiliary Ising variables

Machine learning for phase transition has received intensive research interest in recent years. However, its application in percolation still remains challenging. We propose an auxiliary Ising mapping method for machine learning study of the standard percolation as well as a variety of statistical mechanical systems in correlated percolation representations. We demonstrate that unsupervised machine learning is able to accurately locate the percolation threshold, independent of the spatial dimension of system or the type of phase transition, which can be first order or continuous. Moreover, we show that, by neural network machine learning, auxiliary Ising configurations for different universalities can be classified with high confidence level. Our results indicate that the auxiliary Ising mapping method, despite of it simplicity, can advance the application of machine learning in statistical and condensed-matter physics.

cond-mat.stat-mech↗