Understanding occupants' behaviour, engagement, emotion, and comfort indoors with heterogeneous sensors and wearables

Nan Gao, Max Marschall, Jane Burry, Simon Watkins, Flora D. SalimarXiv 2021 · OpenReview

TL;DR

This paper introduces two novel, publicly available datasets, In-Gauge and En-Gage, collected from a K-12 school, that combine environmental sensor data with wearable physiological and self-reported data from students and teachers. The goal is to foster research into understanding the complex interplay between indoor environments and occupant well-being, engagement, and behavior, paving the way for intelligent building systems.

What Problem Does This Paper Address?

Designing buildings that optimize occupant well-being while minimizing energy consumption is a significant challenge. A key hurdle is the "performance gap" between simulated and actual building performance, largely attributed to the simplistic modeling of occupant behavior. Current models often rely on static rules (e.g., turning on AC at 24°C) that fail to capture the nuanced, probabilistic nature of human actions. Furthermore, understanding the holistic experience of occupants—beyond just thermal comfort—to include engagement and emotional states is crucial for creating truly supportive environments. This paper addresses this by providing rich, multi-modal data to enable more sophisticated modeling of occupant behavior and well-being.

Methods

The study employed a mixed-methods approach combining two distinct data collection efforts:

  • In-Gauge (Longitudinal Study): This 5-month study focused on environmental data. It involved deploying outdoor weather stations and indoor environmental sensors (temperature, humidity, CO2, noise) in 17 classrooms. Crucially, temperature sensors were also placed on air-conditioning vents to infer AC usage. Data was logged at 5-minute intervals, providing a detailed picture of the indoor climate and occupant presence.
  • En-Gage (Cross-sectional Study): This 4-week study focused on occupant data. 23 students and 6 teachers wore Empatica E4 wristbands to collect physiological data (EDA, PPG, ACC, skin temperature). Participants also completed daily online surveys to report on thermal comfort, learning engagement, emotions, and seating locations.

The combination of these datasets allows for the analysis of relationships between environmental conditions and occupant states.

Datasets

The paper introduces two new, publicly available datasets:

  • In-Gauge Dataset: This dataset comprises longitudinal environmental data from a K-12 school, including outdoor weather conditions and indoor climate parameters (temperature, humidity, CO2, noise) from 17 classrooms. It also includes inferred air-conditioning usage. The data is provided in CSV files, one for each classroom, with a 5-minute logging frequency.
  • En-Gage Dataset: This dataset captures cross-sectional data from 23 students and 6 teachers. It includes raw physiological data from Empatica E4 wristbands (ACC, EDA, BVP, HR, TEMP) and self-reported data on thermal comfort, engagement, emotions, and seating locations. The dataset is available in both raw and organized formats, based on class groups.

These datasets are distinct from existing benchmarks by offering a unique combination of detailed environmental monitoring within operational buildings and rich, multi-modal physiological and self-reported data from a specific demographic (high school students and teachers).

Key Insights and Ideas

  • Holistic Occupant Understanding: The core idea is that true occupant well-being and building performance optimization require understanding more than just thermal comfort. Integrating physiological and behavioral data with environmental data is key to capturing the complex interplay of factors influencing how people feel and act in indoor spaces.
  • Bridging the Performance Gap: By providing real-world data on occupant behavior (e.g., AC usage) and environmental conditions, these datasets can help researchers develop more accurate predictive models for building performance simulations, thus reducing the gap between design intent and actual performance.
  • Enabling Intelligent Feedback Systems: The ultimate goal is to use these rich datasets to design intelligent feedback systems that can proactively adjust building environments or provide personalized recommendations to occupants, enhancing both comfort and energy efficiency.
  • The Value of Heterogeneous Sensing: The paper highlights the power of combining diverse sensing modalities—environmental sensors, wearables, and self-reports—to gain a comprehensive understanding of human-building interaction.

Key Results

The paper primarily focuses on introducing and describing the datasets. While specific quantitative results from analyses are not the main focus, the paper demonstrates potential applications:

  • Predictive Modeling Potential: Figures 11 and 12 illustrate how the longitudinal data can be used to fit logistic regression models predicting air-conditioner switching behavior based on indoor and outdoor temperatures.
  • Data Richness: The datasets contain extensive information, including 488 survey responses and over 1400 hours of wearable data for the En-Gage study, and 5.5 months of environmental data for the In-Gauge study.
  • Quality Control: The paper details efforts in data cleaning and quality control for wearable signals, noting an average of 2.66% missing data due to various factors, which was reduced to 2.46% after applying a median filter for EDA data.

Limitations and Future Work

Limitations:

  • AC Usage Inference: The paper acknowledges that directly monitoring air-conditioning unit usage was not possible. Instead, it relied on inferring usage through temperature fluctuations, which is a "crude method" and a potential source of error, requiring manual overwriting of some data frames.
  • Wearable Sensor Accuracy: The accuracy of Empatica E4 wristbands is noted as limited, and they are susceptible to noises from various factors like loose electrodes or faulty wiring.
  • Generalizability: The study was conducted in a single K-12 private school, which may limit the generalizability of findings to other building types or demographics.

Future Work:

  • Predictive Models: The authors suggest that the datasets are ideal for developing predictive models of occupant behavior, particularly for air-conditioning usage.
  • Relationship Analysis: Future research can explore the relationships between indoor/outdoor climates and physiological signals to predict occupant states.
  • Intelligent Feedback Systems: The datasets can be used to design and test intelligent feedback systems for buildings.
  • Applications in Education: The data can inform applications such as monitoring student disengagement, studying peer effects, and optimizing indoor environments for learning.
  • Further Data Validation: The authors suggest that additional measures could be employed to validate wearable data, such as comparing signals from different body parts or using appropriately designed studies.

Summary updated

Abstract

We conducted a field study at a K-12 private school in the suburbs of Melbourne, Australia. The data capture contained two elements: First, a 5-month longitudinal field study In-Gauge using two outdoor weather stations, as well as indoor weather stations in 17 classrooms and temperature sensors on the vents of occupant-controlled room air-conditioners; these were collated into individual datasets for each classroom at a 5-minute logging frequency, including additional data on occupant presence. The dataset was used to derive predictive models of how occupants operate room air-conditioning units. Second, we tracked 23 students and 6 teachers in a 4-week cross-sectional study En-Gage, using wearable sensors to log physiological data, as well as daily surveys to query the occupants' thermal comfort, learning engagement, emotions and seating behaviours. Overall, the combined dataset could be used to analyse the relationships between indoor/outdoor climates and students' behaviours/mental states on campus, which provide opportunities for the future design of intelligent feedback systems to benefit both students and staff.
BibTeX
@misc{gao2021understanding,
  author = {Nan Gao and Max Marschall and Jane Burry and Simon Watkins and Flora D. Salim},
  title = {{Understanding occupants' behaviour, engagement, emotion, and comfort indoors with heterogeneous sensors and wearables}},
  year = {2021},
  eprint = {2105.06637},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2105.06637},
}
Figure