Personalizing Thermal Comfort in Shared Environments

16 concepts12 papers1 proposal

Traditional heating, ventilation, and air conditioning (HVAC) systems operate on a flawed premise: that a single, static temperature setpoint can satisfy everyone in a room. For decades, building management systems have relied on aggregate metrics like the Predicted Mean Vote (PMV) index to calculate an "average" comfort level for a space.

However, as Laura Klein et al. point out in Coordinating occupant behavior for building energy and comfort management using multi-agent systems, these standardized frameworks mask massive variations in individual comfort. Human thermal preference is highly subjective, shaped by a complex interplay of physiology, metabolic rates, clothing choices, and behavioral habits (Emil Laftchiev et al. in Personalizing Individual Comfort in the Group Setting).

When a system forces a uniform environment onto a diverse group, it creates a double penalty: widespread occupant discomfort and massive energy waste from over-conditioning spaces (Abraham Hang-Yat Lam et al. in Carrying My Environment with Me: A Participatory-sensing Approach to Enhance Thermal Comfort). To solve this, a growing body of research shifts the paradigm away from rigid physical modeling toward individualized, data-driven thermal comfort optimization.


Learning the Subjective Profile

The first step in personalizing an environment is moving away from static physics equations and toward learning how individuals actually experience temperature.

Rather than relying on generic standards, modern approaches treat thermal preference as a latent variable to be inferred through data. For example, Frederik Auffenberg et al. utilize a Bayesian Network in A Personalised Thermal Comfort Model Using a Bayesian Network to map out individual user preferences and handle the uncertainty inherent in human feedback.

To make these systems practical without invading user privacy or demanding constant manual input, researchers look to ambient and behavioral cues:


The Multi-Occupant Dilemma

Learning a single person's preference is straightforward; the real challenge arises in shared commercial spaces where multiple occupants have conflicting comfort demands. If one occupant is too hot and another is too cold, a single thermostat zone cannot satisfy both (Di Xie et al. in Distributed Real-Time HVAC Control for Cost-Efficient Commercial Buildings Under Smart Grid Environment).

To resolve these conflicts, researchers have developed three core strategies:

1. Micro-Climate Regulation

Instead of treating a room as a homogenous block of air, systems can target localized zones. Emil Laftchiev et al. combine weakly-supervised neural networks with room thermodynamic autoencoders to control local micro-climates to within 0.5°C of an individual's ideal preference, proving that high-precision localization can bypass the need for a single room-wide compromise.

2. Human-in-the-Loop Recommender Systems

If the environment cannot adapt to the people, the system can coordinate the people within the environment. Peter Wei et al. propose a deep reinforcement learning system called RECA in A multi-task deep reinforcement learning-based recommender system for co-optimizing energy, comfort, and air quality in commercial buildings with humans-in-the-loop. RECA pairs automated thermostat adjustments with real-time room-change recommendations, guiding occupants to spaces that already match their thermal profiles.

3. Democratic Voting and Feedback Loops

Another approach bypasses complex thermal modeling entirely by letting occupants vote on their comfort. Sean Purdon et al. demonstrate in Model-free HVAC control using occupant feedback that smartphone-based voting allows a shared space's climate to "drift" naturally toward outdoor ambient temperatures, intervening only when occupants explicitly signal discomfort. To make these manual adjustments more intuitive, Gaëlle Calvary et al. explore touch-based thermal feedback in Making Thermostats Great Again: Investigating Touch-Based Thermal Sensation for Manual Control of Thermostats, allowing users to feel immediate tactile cues to better anticipate how their adjustments will impact the room.

Traditional HVAC:  [Fixed Setpoint] ──> Forced on All Occupants ──> Discomfort & Waste

Personalized HVAC: [Occupant Profiles] ──> [Negotiation Engine] ──> Localized Micro-climates
                   [Real-Time Feedback] ──> (RL / Bayesian / MAS) ──> Behavioral Recommendations

Why It Works: The Intuition of "Soft Comfort"

It seems counterintuitive that catering to individual whims could actually save building energy. One might assume that satisfying multiple distinct preferences would require running HVAC systems twice as hard.

The reason this approach works comes down to a concept known as soft comfort. In traditional control systems, comfort is treated as a rigid, hard constraint (e.g., the room must be kept at exactly 22°C). In contrast, personalized systems treat comfort as a flexible, negotiable distribution.

In Supply Demand Coordination for Building Energy Saving: Explore the Soft Comfort, Zhanbo Xu demonstrates that humans do not possess a single static comfort point, but rather a range of tolerance that shifts based on context—such as the financial or environmental cost of energy. By modeling heterogeneous occupant archetypes (such as frugal, neutral, or comfort-focused users), building systems can exploit this inherent human elasticity.

When energy prices spike or grid demand peaks, the system can mathematically negotiate with the learned occupant models. It identifies who is willing to tolerate a slight temperature drift, concentrates conditioning resources only on those who genuinely need it, or leverages low-energy alternatives like automated window ventilation (Gabriel Jacoby-Cooper et al.). By transforming comfort from a rigid constraint into a multi-agent optimization problem, personalized climate control unlocks massive, hidden degrees of freedom for building efficiency.

Key Foundational Work

Carrying My Environment with Me: A Participatory-sensing Approach to Enhance Thermal Comfort (2013) — contributed 2 concept elements to this direction.
Model-free HVAC control using occupant feedback (2013) — contributed 1 concept elements to this direction.
Coordinating occupant behavior for building energy and comfort management using multi-agent systems (2011) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_subjective_thermal_preference_b8fb314f,
  title = {subjective thermal preference},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/subjective-thermal-preference/rdr_b8fb314fde864764ba850cf8dd552089},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_subjective_thermal_preference_b8fb314f_v1,
  title = {subjective thermal preference},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/subjective-thermal-preference/rdr_b8fb314fde864764ba850cf8dd552089/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake