Transfer of Human Preference to Smart Machines: A Case Study of Human Thermal Comfort Control

A new approach to thermal comfort control is presented. Thermal comfort control systems are complex systems in which the control objectives are vaguely defined by human preference. It is shown that there is a distict difference in the environmental preference from one person to another based on votes of thermal sensation provided by human subjects. A recursive algorithm for learning the individual preference is proposed. The algorithm changes a prediction of the thermal sensation by locally changing a stored representation of the system.