About
Undergraduate researcher at Tsinghua University working at the intersection of AI and the built environment. My research focuses on mobile robotic sensing for indoor environmental monitoring, graph-based building energy modeling, and vision-language models for occupant-centric building intelligence. I build open datasets and autonomous platforms that bridge data-driven methods with building science.
Education
B.Eng. Interdisciplinary Engineering · 2023 – 2027
Sustainable Cities & Ecological Space, Xiuzhong College, Tsinghua University
Expertise
Research
ARMIS: Autonomous Mobile Sensing System for Indoor Environmental Monitoring
Built an autonomous mobile robot integrating IEQ sensors with multimodal vision. Applied Gaussian Process Regression for high-resolution environmental field reconstruction. Fused YOLOv5 with GPT-4o / GLM-4v VLMs for occupant characterization and thermal comfort assessment. Validated across 7,700 m² at Tsinghua. First / Co-first Author.
ArchEGraph: Geometry–Topology–Physics Aligned Building Energy Graph Dataset
Constructed a benchmark of 5,481 buildings (49,326 cases) as heterogeneous graphs aligning geometry, topology, meteorology, and thermal loads. Defined M2G and G2E tasks with cross-building, cross-climate generalization. Tsinghua × UC Berkeley × MIT. Co-first Author (Equal).
IEQ & Biosafety Mobile Sensing and Early-Warning Platform
Mobile robotic platform for post-pandemic microbial aerosol monitoring. Proposed the MERI (Microbial Equivalent Risk Index) for joint biosafety–comfort assessment. Awarded 3rd Prize at Tsinghua Challenge Cup. First Author.
National Key R&D: Intelligent Pre-Planning & Post-Occupancy Evaluation
Contributed to the 14th Five-Year Plan project (2022YFC3801300) led by Academician Weimin Zhuang. Conducted on-site environmental monitoring at Dahe Village Archaeological Museum. Research Assistant.
MobilEnv: Multi-Scene, Multi-Modal Indoor Environmental Dataset
Open-source dataset covering 50,000+ m² across multiple climate zones — including Beijing Daxing Airport (27,000 m²) and Dahe Village Museum. CO₂, PM₂.₅, temp, humidity, illuminance, robot trajectories, reconstructed fields. First Author. GitHub
Building Environment Perception & Reconstruction
Invited talk on AI-driven environmental perception and spatial reconstruction at the Architectural Society of China Student Chapter Annual Conference.
Publications
ARMIS: An Autonomous Mobile Sensing System for Intelligent Indoor Environmental Monitoring with Multimodal Data
Indoor Air 2026 — Oral Presentation
First / Co-first Author
ArchEGraph: A Large-Scale Graph Dataset for Geometry–Topology–Physics Aligned Building Energy Modeling
Submitted to NeurIPS 2026 Datasets & Benchmarks Track
Co-first Author (Equal Contribution) · Tsinghua × UC Berkeley × MIT