Physical​ and physiological information is essential for indoor human monitoring applications. In order to capture such information non-intrusively, we utilize/design sensors that can capture the vibration induced by people/body parts to infer a variance of activities (over different scales, both infrastructure and wearable) in the physical world. We explore new opportunities to utilize complementary information over different scales (on-body, room, city, etc.) to achieve accurate inference.

Multiple co-located sensing modalities capture the same physical event from different aspects and provide complementary information for inference. However, to accurately associate signals from different modality capturing the same event is challenging, expecially when the modalities are significantly different.

Data acquisition quality directly affects the information representative and the model accuracy in applications of real-world deployed cyber-physical systems. We combine physical and data-driven knowledge to design metrics and methods to assess the signal and dataset quality for particular sensing tasks.

The heterogeneity of cyber-physical systems brings challenges and opportunities for various applications. Our goal is to explore new methods to combine the complementary characteristics of multiple sensing modalities for accurate fine-grained learning with limited (labeled) data.

San Joaquin Valley's (SJV) residents face challenges of air pollutant exposure -- the basin terrain of the SJV traps air pollutants easily. At the same time, the local land use (e.g., industry, agriculture) results in a high variation of pollutant sources of PM2.5. We develop Artificial Intelligence of Things (AIoT) with both advanced sensing and learning algorithms to tackle the challenge with low-cost and scalable solutions.

Latest Papers:

Shangjie Du, Zhizhang Hu, and Shijia Pan. 2024. GraPhy: Graph-Based Physics-Guided Urban Air Quality Modeling for Monitoring-Constrained Regions. In Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation (BuildSys ’24), November 7–8, 2024, Hangzhou, China.

Zhizhang Hu∗, Shangjie Du∗, Yuning Chen, Xuan Zhang, Wan Du, Asa Bradman, and Shijia Pan. 2023. Poster Abstract: Enhancing Fault Resilience of Air Quality Monitoring in San Joaquin Valley: A Data Equity Analysis. In Proceedings of the 21st ACM Conference on Embedded Networked Sensor Systems (SenSys ’23), November 12–17, 2023, Istanbul, Turkiye. ACM, New York, NY, USA, 2 pages.

Wang, Gang, Shijia Pan, and Susu Xu. "Decoupling the unfairness propagation chain in crowd sensing and learning systems for spatio-temporal urban monitoring." In Proceedings of the 8th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, pp. 200-203. 2021.

Award: Best Paper Award Runner-Up BuildSys'24

California’s San Joaquin Valley (SJV) is the key contributor to the nation’s agriculture production, with 7984 square miles of prime farmland, farmland of statewide importance, or unique farmland. However, the region faces escalating climate changes with frequent events such as wildfires and droughts. These severe climate events directly affect agricultural yields, posing financial challenges to the already disadvantaged communities. A high-granularity and high-resolution multimodal (e.g., soil moisture, temperature, etc.) map of the farmland is essential for enabling precision agriculture control (e.g., irrigation ) and enhancing the farm’s climate resilience.

Latest Paper:
Yue Zhang, Abdias Tellez Benitez, Reza Ehsani, and Shijia Pan. "DaQual: Data Quality Assessment for Tree Trunk Relative Water Content Sensors in a Pomegranate Orchard." In Proceedings of the 20th ACM Conference on Embedded Networked Sensor Systems, pp. 1096-1101. 2022.