A Mobile robot based personalized thermal comfort scheme

Abstract

This study designs a robot control system to improve the accuracy of recognizing human states related to thermal comfort in real indoor spaces. The system aims to accurately estimate thermal comfort through vision-based recognition while minimizing the robot's movement to relieve user disturbance by robot's movement. It dynamically navigates to optimal positions by evaluating the confidence of vision-based human state recognition. The control system trains control policy by Deep Reinforcement Learning. The real-world evaluation demonstrates the system's effectiveness in accurately recognizing human states with minimal movement trajectories. This robot-based system can contribute to personalized thermal control for estimating accurate thermal comfort.