International Workshop on Environmental Sensing Systems for Smart Cities

Program

In 2025, we are pleased to announce that the EnvSys Workshop will once again be merged with the NetAISys Workshop, forming a combined, day-long event. While the two workshops focus on different yet complementary aspects of AI, they share a common vision: integrating AI into real-world systems. EnvSys emphasizes applying AI technologies to environmental sensing—addressing pressing environmental issues through tech-driven solutions—whereas NetAISys explores the use of AI in networking systems, investigating the challenges and opportunities it presents. We believe this merger will encourage vibrant discussions and promote the exchange of ideas across a broader spectrum of research at the intersection of networking, AI, and environmental sensing. The focus areas of EnvSys, in particular, provide compelling real-world use cases for networked AI systems. The joint workshop program will feature eleven papers, including five from the EnvSys workshop and six from the NetAISys workshop. Each paper presentation is allocated 15 minutes, followed by a 3-minute Q&A session. The event will also include two keynotes and a panel discussion to foster interactive and cross-disciplinary dialogue.

---

• 08:00 – 09:00: Registration and Breakfast
• 09:00 – 09:05: Opening Remarks

---

• 09:05 – 09:45: Keynote 1 – Speaker: Prof. Yasamin Mostofi (University of California Santa Barbara)
• Title: Environmental Sensing and Perception with Wireless Signals
• Abstract: Communication signals are ubiquitous these days. This has inspired using them beyond communication, e.g., for sensing and learning about the environment. In this talk, I will provide an overview of some of our work on mathematical modeling, design principles, and practical applications for wireless sensing and perception. I will start by examining crowd analytics, with an emphasis on understanding collective behaviors and uncovering emerging crowd patterns. Along this line, I will present a new mathematical modeling framework, using tools such as stochastic geometry and queuing theory, that captures and extracts key collective crowd parameters. Next, I will focus on scene understanding. I will set forth that the scattered RF signals off of objects carry rich information about the edges of the objects. Based on this observation, I then propose a new way of thinking about RF imaging and scene understanding, via edge tracing. More specifically, I will show how the Geometrical Theory of Diffraction (GTD) and the corresponding Keller cones can be exploited to image edges of the objects. I will then demonstrate the applicability of this approach by showing how WiFi can image and read the English alphabets through walls. In the third part of the talk, I will then address one major issue in applying deep learning to RF sensing problems: lack of large enough RF training data to achieve generalizable results. Along this line, I will show how recent advances in computer vision can be harnessed to develop generalizable machine learning models for RF sensing, and further discuss how this approach has enabled a large, successful clinical trial for gait disorder assessment using commodity WiFi signals. Finally, I will discuss how unmanned vehicles can expand the capabilities of wireless sensing systems.
• Biography: Yasamin Mostofi received the B.S. degree in electrical engineering from Sharif University of Technology, and the M.S. and Ph.D. degrees from Stanford University. She is currently a professor in the Department of Electrical and Computer Engineering at the University of California Santa Barbara. Yasamin is the recipient of the Presidential Early Career Award for Scientists and Engineers (PECASE), the Antonio Ruberti Prize from the IEEE Control Systems Society (research contribution award for 40 and under), the National Science Foundation (NSF) CAREER award, and the IEEE Outstanding Engineer Award of Region 6 (more than 10 Western U.S. states), among other awards. She is a fellow of IEEE. She was a semi-plenary speaker at the 2018 IEEE Conference on Decision and Control (CDC) and a keynote speaker at the 2018 Mediterranean Conference on Control and Automation (MED). Yasamin's research is multi-disciplinary, in the two areas of wireless systems and robotics. Current high-level research thrusts include 1) RF sensing for several different applications such as occupancy analytics, collective behaviors, through-wall imaging, context inference, smart health, and smart spaces; and 2) communication-aware robotics, UAV-assisted connectivity, and joint robotic path planning and communication. Her research has appeared in several reputable news venues such as BBC, New Scientist, Daily Mail, Engadget, TechCrunch, NSF Science360, ACM News, and IEEE Spectrum, among others. Yasamin has served in many different professional capacities over the years. Recent samples include serving on the inaugural editorial board of NPJ Wireless Technology as part of Nature Portfolio, serving on the Board of Governors of IEEE CSS, serving as a senior editor for IEEE TCNS, and serving as a program co-chair for ACM MobiCom 2022, among others.

---

• 09:45 – 10:55: Session 1 – Edge AI and Intelligent Sensing Systems
• Bringing Edge Intelligence to Wildlife Camera Traps with Hyperdimensional Computing (NetAISys)
  Jida Zhang, Timothy Jacques, Joseph Chen, Zerina Kapetanovic (Stanford University)
• Sensor‑free Microclimate Monitoring Using Existing LoRaWAN Signal Characteristics (EnvSys)
  Fateme Nikseresht, Victor Ariel Leal Sobral, Moeen Mostafavi, Brad Campbell (University of Virginia)
• Wi‑Chat: Large Language Model‑powered Wi‑Fi‑based Human Activity Recognition (EnvSys)
  Yili Ren (University of South Florida); Haopeng Zhang, Haohan Yuan (University of Hawaii at Manoa); Jingzhe Zhang, Yitong Shen (University of South Florida)
• Presto: Hybrid CPU‑GPU Preprocessing Framework for Video‑based AI Inference System (NetAISys)
  Jihyuk Lee (Chung‑Ang University); Dongsu Han (KAIST); Jaehong Kim (Carnegie Mellon University)

---

• 10:55 – 11:00: Break

---

• 11:00 – 11:55: Session 2 – ML Pipelines, Metadata, and Infrastructure
• MRM3: Machine Readable ML Model Metadata (NetAISys)
  Andrej Čop, Blaž Bertalanič, Marko Grobelnik, Carolina Fortuna (Jožef Stefan Institute)
• SensorMCP: A Model Context Protocol Server for Custom Sensor Tool Creation (NetAISys)
  Yunqi Guo, Guanyu Zhu, Kaiwei Liu, Guoliang Xing (The Chinese University of Hong Kong)
• Empirical Analysis of LLMDPP: Advancing Log Parsing in the LLM Era (NetAISys)
  Kehan Wang, Siqin Zhang, Haijing Nan, Xueyu Hou (China Telecom Cloud Computing Corporation); Jiaqi Zou (Tsinghua University); Zicong Miao (China Telecom Cloud Computing Corporation)

---

• 11:55 – 13:00: Lunch Break

---

• 13:00 – 14:00: Panel Discussion
• Panelists: Prof. Lin Zhong (Yale CS), Prof. Hang Qiu (UC Riverside ECE/CSE), Prof. Juheon Yi (Microsoft), and Prof. Yasamin Mostofi (UCSB)

---

• 14:00 – 14:40: Keynote 2 – Speaker: Juheon Yi, Senior Researcher (Microsoft)
• Title: Edge‑Cloud Cooperative Platform for Interactive Video Analytics
• Biography: Juheon Yi is a senior researcher at Microsoft Research Asia. His research interests lie in edge AI systems and video analytics. Specifically, his research focuses on characterizing the workloads of interactive video analytics applications and building core mobile/edge systems to support them. Juheon’s work has been consistently recognized at top-tier conferences and journals, including ACM MobiCom, MobiSys, Multimedia, IEEE INFOCOM, and IEEE Transactions on Mobile Computing. He is a recipient of the Microsoft Research PhD Fellowship, Best Paper Award in ACM Students in MobiSys 2021 Workshop, and Best PhD Dissertation Award from the Department of Computer Science and Engineering (CSE), Seoul National University (SNU).

---

• 14:40 – 15:05: Coffee Break

---

• 15:05 – 16:20: Session 3 – Intelligent Environments & Context‑Aware AI (Part 3)
• SPATIUM: A Context‑Aware Machine Learning Framework for Immersive Spatiotemporal Health Understanding (NetAISys)
  Yang Liu, SiYoung Jang, Alessandro Montanari, Fahim Kawsar (Nokia Bell Labs)
• Localization using Angle‑of‑Arrival (AoA) Triangulation (EnvSys)
  Amod K. Agrawal (Amazon Lab)
• Adaptive Water pH Sensing in Variable Conditions Using Near Infrared Imaging and Machine Learning (EnvSys)
  Fadoua Khmaissia, Nirupama Ravi (Nokia Bell Labs)
• CrashSniffer: UWB‑Based Anchor‑Free Pedestrian Collision Prediction for Personal Mobility Vehicles (EnvSys)
  Taeckyung Lee (KAIST); Juseung Lee (Korea University); Ryuhaerang Choi, Seungjoo Lee, Hyeongheon Cha, Hyungjun Yoon, Song Min Kim (KAIST); Sangwook Bak (Samsung Electronics); Sung‑Ju Lee (KAIST)

---

• 16:20 – 16:25: Closing Remarks