I am a postdoc at the Smart Centre, Massachusetts Institute of Technology (MIT), working with Prof. Jinhua Zhao. I was a Research Fellow in the Department of the Built Environment, National University of Singapore (NUS), where I worked with Prof. Yangmiang Goh. I received my Ph.D. degree in the Department of Automation, Tsinghua University (THU) in 2023. As a member of CFINS, I’m supervised by Prof. Qianchuan Zhao.

My research interest lies in smart buildings and smart cities, including building occupancy detection/estimation, computer vision and generative models. I have published more than 15 first-author papers in top international journals, including Cell Patterns, Building and Environment, Automation in Construction, Energy and Buildings, etc.

πŸ”₯ News

πŸ“– Educations

  • 2024 - present, Postdoctoral Associate, Smart centre, Massachusetts Institute of Technology (MIT). Prof. Jinhua Zhao.
  • 2023 - 2024, Research Fellow, Department of the Built Environment, National University of Singapore. Prof. Yang Miang Goh
  • 2018- 2023, Ph.D., Department of Automation, Tsinghua University. Prof. Qianchuan Zhao
  • 2022, visiting student, Electrical & Computer Engineering, University of Illinois Urbana-Champaign. Prof. Tamer Basar
  • 2015 - 2018, Bachelor, Ye Peida lab, Beijing University of Posts and Telecommunications (BUPT).
  • 2014 - 2018, Bachelor, Department of Automation, Beijing University of Posts and Telecommunications (BUPT).

πŸ“ Publications

1:Co-first authors

*:Corresponding authors

Under-review Paper

sym
  • Toward Pedestrian Head Tracking: A Benchmark Dataset and an Information Fusion Network. Kailai Sun, Xinwei Wang, Shaobo Liu*, Qianchuan Zhao*, Gao Huang, Chang Liu.
  • We collect and annotate a Chinese large-scale cross-scene dataset (Cchead) for crowded head tracking, including 10 diverse scenes of 50,528 frames with over 2,366,249 heads annotated.
  • We present a novel solution and insight for head tracking by leveraging pseudo-multi-source information fusion, achieving SOTA performance.
sym
sym
  • Topology-Aware Hypergraph Reinforcement Learning for Indoor Occupant-Centric HVAC Control. Dianyu Zhong, Tian Xing, Kailai Sun*, Ziyou Zhang, Qianchuan Zhao and Jian Kang. (Corresponding author)
  • Exploring large language models for indoor occupancy detection and estimation for smart buildings. Irfan Qaisar1, Kailai Sun1,*, Qianchuan Zhao. (Corresponding and Co-first author)

  • A computer vision-based construction housekeeping monitoring system: A two-stage object detection-then-classification approach. Zherui Shao, Kailai Sun, Yang Miang Goh*, Jing Tian, Vincent J.L. Gan.

  • Dynamic Occupancy Measurement for Smart Buildings: A Few-shot Large Language Model Approach. Irfan Qaisar, Kailai Sun*, Ziyou Zhang, Qianchuan Zhao. (Corresponding author)

Journal Paper

ENB 2020
sym
  • A review of building occupancy measurement systems.
    Kailai Sun, Qianchuan Zhao*, Jianhong Zou. Energy and Buildings (ENB). 216 (2020): 109965. (Top , SCI, JCR Q1, IF:7.093, citations:163, ESI highly cited paper). (Acceptance Rate: 22%)
  • Five-years occupancy measurement systems based on cameras, WiFi, PIR sensors, CO2 sensors, electricity sensors are analyzed and discussed.
  • The first comprehensive analysis based on different types and installed locations of camera.
Cell Patterns 2022
sym
  • Honeycomb: An open-source distributed system for smart buildings. Tian Xing1, Hu Yan1, Kailai Sun1, Yifan Wang, Xuetao Wang and Qianchuan Zhao*. Patterns.(Cell sub journal)(SCI, JCR Q1, IF:6.5 )(2022). Data link. Code link.
  • A bee-inspired, fully distributed, and open-source building IoT solution, has strong flexibility and robustness, multiple functionalities.
  • Proposed vision-based deep-learning occupancy measurement system.
  • High user acceptance during long-term stable operation.
2025
sym
  • A review of AI edge devices and lightweight CNN and LLM deployment. Kailai Sun1,*, Xinwei Wang1, Qianchuan Zhao. Neurocomputing. (2025). (SCI, JCR Q1, Top, IF:5.5).
  • Many AI edge devices with AI accelerators (Nvidia Nano, TX2, Intel NCS2, Google Coral Dev Board, Coral USB, Baidu Edgeboard FZ3, AMD Kria K2, etc) are compared and analyzed.
  • Lightweight CNN and large language models, neural network compression, and their deployment on AI edge devices for practical applications, are reviewed.
BAE 2022
sym
  • A Fusion Framework of Vision-based Indoor Occupancy Estimation.
    Kailai Sun1, Peng Liu, Tian Xing, Qianchuan Zhao* and Xinwei Wang. Building and Environment (BAE). 222 (2022): 109354.(Top , SCI, JCR Q1, IF:7.201). Code link. (Acceptance Rate: 17%)
  • The first work to develop a three-level fusion framework in vision-based indoor occupancy estimation.
  • Our framework achieves state-of-the-art (SOTA) performance through ablation studies on practical building surveillance videos.
2024
sym
  • Change Detection Network for Construction Housekeeping using Feature Fusion and Large Vision Models. Kailai Sun, Zherui Shao, Yang Miang Goh*, Jing Tian, Vincent J.L. Gan. Code link. Automation in Construction.(AIC). 2025. (Top , SCI, JCR Q1, IF:9.6).(Acceptance Rate: 14%)
  • We propose a change detection dataset to mitigate housekeeping problems in construction sites.
  • We designed a novel Housekeeping Change Detection Network (HCDN) with Large Vision Model (LVM) and more effective features fusion, achieving superior performance than existing SOTA methods.
  • We integrate HCDN into the proposed housekeeping detection system, which shows its potential in practical construction site situations.
BAE 2022
sym
  • MPSN: Motion-aware Pseudo-Siamese Network for Indoor Video Head Detection in Buildings. Kailai Sun, Xiaoteng Ma, Peng Liu, Qianchuan Zhao*. Building and Environment (BAE). 222 (2022): 109354.(Top , SCI, JCR Q1, IF:7.201).Code link. (Acceptance Rate: 17%)
  • The first work to jointly train the current frame and the pixel-level motion information into an end-to-end CNN network in head detection.
  • We validate its robustness through adversarial experiments with a mathematical solution of small perturbations.
BAE 2024
sym
  • DMFF: Deep Multimodel Feature Fusion for Building Occupancy Detection. Kailai Sun*. Building and Environment (BAE). (2024).(Top , SCI, JCR Q1, IF:7.201). Invited. Code link.
  • Multi-model deep learning meets occupancy. Sound, image and environmental data are integrated to detect occupancy.
  • To transfer pre-trained AI models and domain knowledge into the building field, We introduce the pretrain-finetune pipeline and propose a Transformer-based multimodel fusion algorithm for occupancy detection.
BAE 2022
sym
  • Building occupancy number prediction: A Transformer approach. Kailai Sun1,*, Irfan Qaisar1,Muhammad Arslan Khan, Qianchuan Zhao. Building and Environment (BAE). (2023).(Top , SCI, JCR Q1, IF:7.201). Code link. Accepted with a first-round Minor revision. (Acceptance Rate: 17%)
  • We publicly provide an actual building operating dataset for six zones.
  • Our Transformer-based occupancy prediction algorithm performs superior on our dataset than other existing deep learning algorithms
ENB 2024
sym
  • High-accuracy occupancy counting at crowded entrances for smart buildings. Kailai Sun, Xingwei Wang, Tian Xing, Shaobo Liu*, Qianchuan Zhao. Energy and Buildings (ENB). (2024).(Top , SCI, JCR Q1, IF:7.093). (Acceptance Rate: 22%) Code link.
  • The advanced computer vision model for sensing and tracking occupancy heads using surveillance videos.
  • The practical occupant-centric control (OCC) for energy efficiency and occupant comfort.
  • This research can be generalised to border domains, including building management, flow estimation, safety prevention, etc.
AAP 2023
sym
BAE 2025
sym

Conference Paper

  • Multi-Source Information Fusion network for building occupancy estimation.Kailai Sun, Tian Xing, Xinwei Wang, Zhou Yang, and Qianchuan Zhao*. In: Alicja Maciejko (eds) Human Factors in Architecture, Sustainable Urban Planning and Infrastructure. AHFE (2023) International Conference. AHFE Open Access, vol 89. AHFE International, USA. http://doi.org/10.54941/ahfe1003687.(Acceptance rate:34%)
sym
sym

✈ Internships

  • 2021.06 - 2021.08, Harbin Institute of Technology Robotics Research Institute Co., Ltd, Yangzhou.

πŸ§‘β€πŸ’» Workshops

πŸ‘¨πŸ½β€πŸ€β€πŸ‘¨πŸΌ Collaborators

  • Qianchuan Zhao - Professor, Department of Automation, Tsinghua University.
  • Yang Miang Goh-Associate Professor, Department of the Built Environment, National University of Singapore.
  • Jian Kang - Professor, Bartlett School Env, Energy & Resources, University College London.
  • Gao Huang-Associate Professor, Department of Automation, Tsinghua University.
  • Jinhua Zhao- Professor, Department of Civil and Environmental Engineering, Massachusetts Institute of Technology.
  • Shaobo Liu - Associate Professor, Intelligent Transportation Systems Research Center, Wuhan University of Technology.
  • Peter Luh - Professor, Board of Trustees Distinguished Professor Emeritus, University of Connecticut.
  • Yueng-hsiang Huang-Associate Professor, Oregon Institute of Occupational Health Sciences, Oregon Health and Sciences University.
  • Xiaoteng Ma - Postdoc, Department of Automation, Tsinghua University.
  • Anbang Liu - Ph.D., Department of Automation, Tsinghua University.

πŸŽ– Honors and Awards

Scholarship

  • 2022.12 The first prize of Tsinghua University Weimin Zheng Scholarship (Top 1%)
  • 2022.11 Tsinghua University - Lingjun Pilot Scholarship
  • 2021.11 Tsinghua Scholarship for Overseas Graduate Studies
  • 2019.10 The first prize of Tsinghua University Comprehensive Scholarship (Top 1%)
  • 2015.06 & 2016.06 Haohan company Scholarship (Top 1%)
  • 2015.09 & 2016.09 & 2017.09 National Encouragement Scholarship (Top 3%)

Student

  • 2023.06 Outstanding graduates of Department of Automation Tsinghua University (Top 17%)
  • 2018.06 Outstanding graduates of Beijing ordinary colleges and universities (Top 1%)
  • 2023.06 Academic Excellence Award in System Engineering Institution, Department of Automation, Tsinghua University
  • 2018.06 Excellent Student Cadre (Top 3%)
  • 2015.09 & 2016.09 & 2017.09 Merit Student (Top 3%)

Competition

  • 2017 American Undergraduate Mathematical Contest in Modeling Honorable Mention
  • 2015.11 The second prize of the Chinese National Undergraduate Mathematical Contest
  • 2016.06 The first prize in the Beijing Division of the Chinese National Undergraduate Mathematical Contest in Modeling
  • 2016.07 The second prize in the Beijing Division of the National Undergraduate Electronic Design Competition
  • 2022 Best Researcher Award for the contribution and honourable achievement in innovative research. International Research Awards on New Science Inventions NESIN 2022 Awards.

πŸ’» Projects

  • Modeling and Intelligent Design Methods for Hydrogen-Inclusive Multi-Energy Supply and Demand Systems (National Natural Science Foundation Major Project 62192751)
  • Performance Optimization and Control of Networked Dynamic Systems (National Natural Science Foundation Major Project 61425027)
  • New generation intelligent building platform (National Key Research and Development Project of China under Grant 2017YFC0704100)
  • Energy Internet Planning, Operation and Trading (National Key Research and Development Project of China under Grant 2016YFB0901900)
  • Research on Pedestrian β€œState-action” Mapping Laws for Subway Station pedestrian Flow Simulation and Inverse Reinforcement Learning Modeling (National Natural Science Foundation General Program 52172308)
  • Satellite remote sensing image interpretation technology and system based on artificial intelligence (2019 Major Science and Technology Program for the Strategic Emerging Industries of Fuzhou, Grant BNR2019TD01009)
  • The 111 International Collaboration Program of China under Grant No. BP2018006
  • The National Innovation Center of High-Speed Train R&D project (CX/KJ-2020-0006)
  • Huawei company-Control security issue literature research project
  • Ridar-based road vehicle counting and intelligent video surveillance system project
  • Intelligent detection and control aircraft system from Tongfang Intelligent Technology Co., ltd
  • Evaluation of (Singapore Changi Airport Group and Hua Wei Singapore Company) CAG’s safety model for profiling construction contractors’ safety performance; Ref No.: 2023-0492
  • Computer Vision for Housekeeping on Construction Site; Award No.: AISG2-100E-2021-084
  • Integrating Climate and Economics Models to Forecast Workplace Safety and Health for Singapore Construction Industry: A Hybrid Econometrics & Machine Learning Approach, FASS-CDE Joint Seed Grant Call (MOE-Tier 1), NUS
  • Mens, Manus and Machina (M3S) – How AI Empowers People, Institutions and the City. National Research Foundation.

βœ’οΈ Reviewer

Journal: IEEE Transactions on Control Systems Technology, IEEE Internet of Things Journal, Building and Environment, Results in Control and Optimizationm, Scientific Programming, Scientific Reports, The Journal of Supercomputing, Journal of Process Mechanical Engineering, etc.

Conference:IEEE International Conference on Robotics and Automation, The World Congress of the International Federation of Automatic Control, etc.

πŸ“š Book

Participated in the translation of the book: Bertsekas, Dimitri. Convex optimization algorithms. Athena Scientific. 2015.

πŸ‘¨β€πŸ« Teaching

Teaching assistant for β€œautomatic control theory” for 8 times (4 years).

πŸ“ž Contact

E-mail: skl.2018@tsinghua.org.cn  18813126518@163.com  skl24@mit.edu   kailai.sun@smart.mit.edu .

πŸ‘¨β€πŸ« Co-supervised students at Tsinghua:

Irfan Qaisar (2022-;Ph.D.)
Dianyu Zhong (2023-;Ph.D.)
Muhammad Arslan Khan (2023-;Ph.D.)
Xinwei Wang (2020-;Master; Gewu Tech. company, Xiaomi Company)
Ruoyu Wang (2023-; Master)
Peng Liu (2021-2023; Master)
Zhou Yang (2021-2023; Master; BYD)
Guanyi Wang (2023-;Master; Huawei)
Zhe Yu (2022-2023; Master)
Zaixian Han (2020; Bachelor; Korea)
Meiling Piao (2019; Bachelor; Korea)
Xi Miao (2023-;Bachelor; U.S)