Personalized Thermal Comfort System Using Reinforcement Learning: Integrating Multimodal Sensing and Decision-Making

Summary:This paper introduces a personalized thermal comfort system using reinforcement learning, which integrates multimodal physiological and environmental sensing data to predict and dynamically adjust individual thermal preferences. The system employs a two-stage approach: first, it builds a user thermal comfort model using sensor data; second, it applies reinforcement learning algorithms for real-time decision optimization to adapt to each user's physiological characteristics and environmental chan

---

Background and Motivation

In smart buildings and HVAC (Heating, Ventilation, and Air Conditioning) systems, personalized thermal comfort is crucial for enhancing user satisfaction and energy efficiency. However, traditional HVAC systems rely on static setpoints and population-level comfort models, which fail to capture individual physiological differences, resulting in suboptimal user experiences and energy waste.

Technical Innovations

1. Multimodal Data Fusion: The system integrates data from physiological sensors (e.g., heart rate, body temperature) and environmental sensors (e.g., temperature, humidity) to build a comprehensive user thermal comfort model.
2. Reinforcement Learning Decision Optimization: It employs reinforcement learning algorithms to perform dynamic decision optimization based on real-time data and environmental changes, providing personalized thermal comfort adjustments.
3. Two-Stage Approach: The system first builds a user thermal comfort model and then applies the model for real-time decision optimization, ensuring high efficiency and adaptability.

Experiments and Results

The research validated the system's effectiveness through simulations and real-world experiments. The results show that, compared to traditional methods, the system can more accurately predict user thermal comfort needs and provide better adjustment schemes in different environments, thereby enhancing user comfort and reducing energy consumption.

Industry Impact and Future Directions

This study offers a new approach to personalized control in smart buildings and HVAC systems, with broad application prospects. As sensor technologies and AI algorithms continue to advance, the system is expected to achieve more efficient and intelligent thermal comfort management in a wider range of scenarios.

Developer Recommendations

• Data Acquisition and Processing: Developers should focus on the acquisition and processing of multimodal data to ensure the accuracy and robustness of the model.
• Reinforcement Learning Algorithm Selection: Choose appropriate reinforcement learning algorithms based on specific application scenarios to achieve optimal performance and efficiency.
• System Integration and Optimization: Consider real-time performance and resource consumption during system integration to ensure feasibility in practical applications.

---

Source: ArXiv Machine Learning (cs.LG) (2026-08-24)

— END —

Tags: #Reinforcement Learning #Multimodal Sensing #Smart Buildings #HVAC #Personalized Control