2026 ASEE Annual Conference & Exposition

Advancements in machine learning and the increase in use of sensors have made it possible to collect data from buildings. The Improvements in Internet of Things (IoT) enable us to monitor indoor environments and analyze sensor data in various ways [3]. The Building Internet of Things (BIoT) has been very useful in reducing energy consumption and costs. Building occupancy aids in the improvement of the energy management systems, allowing the reduction of energy consumption while maintaining occupant comfort [3].   

Building occupancy estimation is predicting how many people are in a room or a building at a specific time [2,6]. It is useful for improving safety, security, energy efficiency, evacuation planning, and occupant comfort [3,4].  In this project we will estimate the spatial occupancy using machine learning, aiming to predict how many people are present in several rooms at the same time and shows their temporal and spatial relationships using environmental sensor data such as HVAC, plug loads, fans, light, and weather conditions to find patterns without using invasive technologies to avoid privacy concerns.  

In this work, as a part of undergraduate summer research, we will be using transformer neural networks which are good at analyzing time-series data and learning complex patterns in multiple spaces [5]. We will implement encoder-only transformer models. Temporal transformer encoders will be applied to each room separately to learn time-based patterns and a shared Spatial Transformer Encoder to get dependencies between the rooms. 

In this project, we will create the encoder model and implement on ROBOD dataset [1] which contains the environmental sensor data and the ground truth. We will then perform experiments to test the effectiveness and verify the accuracy of the model.

References:

[1] Z. D. Tekler, E. Ono, Y. Peng, S. Zhan, B. Lasternas, and A. Chong, “ROBOD, room-level occupancy and building operation dataset,” Building Simulation, vol. 15, no. 12, pp. 2127–2137, Dec. 2022, doi: 10.1007/s12273-022-0925-9 

[2] W. Wang, J. Chen, and T. Hong, “Occupancy prediction through machine learning and data fusion of environmental sensing and WiFi sensing in buildings,” Automation in Construction, vol. 94, pp. 104–114, Oct. 2018, doi: 10.1016/j.autcon.2018.07.007  

[3] A. N. Sayed, Y. Himeur, and F. Bensaali, “Deep and transfer learning for building occupancy detection: A review and comparative analysis,” Engineering Applications of Artificial Intelligence, vol. 115, p. 105254, Oct. 2022, doi: 10.1016/j.engappai.2022.105254

[4] Z. Chen, C. Jiang, and L. Xie, “Building occupancy estimation and detection: A review,” Energy and Buildings, vol. 169, pp. 260–270, Jun. 2018, doi: 10.1016/j.enbuild.2018.03.084

[5] I. Qaisar, K. Sun, Q. Zhao, T. Xing, and H. Yan, “Multi-sensor-based occupancy prediction in a multi-zone office building with Transformer,” Buildings, vol. 13, no. 8, p. 2002, Aug. 2023, doi: 10.3390/buildings13082002

[6] [7] B. Yang, F. Haghighat, B. C. M. Fung, and K. Panchabikesan, “Season-based occupancy prediction in residential buildings using machine learning models,” e-Prime – Advances in Electrical Engineering, Electronics and Energy, vol. 1, p. 100003, 2021, doi: 10.1016/j.prime.2021.100003