PhD Forum Abstract: The Effects of the Thermal Environment on Sleep and Novel Control Solutions
This PhD Forum presentation outlined a dissertation on improving sleep in Singapore’s hot-humid residential context through technological and behavioural solutions. Situated within Project HEATS, the work focuses on how indoor thermal conditions affect sleep and on developing occupant-centric control approaches for real homes. The talk organized the research around two field studies already underway, with a third phase on novel personal comfort modeling still in development.
The first study applies Just-in-Time Adaptive Interventions (JITAIs) to the built environment. Originally developed in mobile health to support self-management, JITAIs are context-aware recommendation systems that deliver timely, personalized support. This study asked whether smartwatch-based nudges can prompt occupants to adjust indoor settings, such as changing the air-conditioning setpoint, adjusting fan speed, opening a window, or avoiding light and noise. The system was implemented on Apple Watch using Cozie, combined with indoor environmental monitoring and ecological momentary assessment, and evaluated through a 60-day micro-randomized trial.
Preliminary results showed that JITAI and non-JITAI days were approximately balanced, with an overall adherence rate of around 20–24%. JITAIs were particularly effective in encouraging changes to AC setpoints and fan speeds, and about half of the adherence nights showed a positive environment-related impact. Adherence was higher among poor sleepers and more agreeable participants. Next steps include analyses of 12 JITAIs covering sleep health, physical activity, and indoor environmental behaviors, together with a planned dataset release.
The second study examined whether laboratory findings on dynamic bedroom temperature control can transfer to real homes. Lab-based work suggests that aligning bedroom air temperature with core body temperature during sleep can improve sleep outcomes, but real bedrooms differ in thermal dynamics, control precision, and occupant behavior. A research prototype was built to collect wearable, environmental, and subjective data, execute remote AC control via Sensibo Sky, and deliver a dynamic setpoint curve: slightly warm near bedtime to reduce sleep onset latency, slightly cool through mid-sleep to support deep sleep, and back to neutral for REM. The protocol was evaluated in Singapore with 18 participants over 30 nights per condition, using nightly micro-randomization between the dynamic profile and a constant neutral baseline.
The desired temperature profile was successfully achieved in around 80% of participants’ bedrooms. The proposed curve did not improve sleep in this deployment, possibly because the intervention was not strong enough. A related pilot nevertheless showed that the same platform can support field deployment of reinforcement learning and learn from occupant interactions and feedback. The talk concluded that the prototype provides a scalable research tool for residential human–building interaction studies, with support for RL customization and automated field operation.
Slide Download: Access the presentation slides here.
Related publication: The Effects of the Thermal Environment on Sleep and Novel Control Solutions