Current smart buildings are based on scenarios, so they are not prepared for unexpected events. We focus our attention on high adaptability of living matters to environmental changes. "Biofication of Living Spaces" is the concept of creating pleasant living environments using this high adaptability. Biofied room is integrated with sensor agent robots to interact with residents and acquire environmental information. In this research, we propose a highly adaptive algorithm to control the devices automatically. Based on physiological adaption, we can make the algorithm very flexible. As the first step in this research, a prototype of the sensor agent robot is built. Camera, microphone, proximity sensor, laser range-finder are mounted on the robot. As a sensor agent robot follows the residents, it acquires environmental information, and records the interaction between the robot and human. In a suggested control model, a resident is built in the control loop and his/her uncomfortable feeling plays a role of control signal. Following its signal, devices are controlled. Results obtained from the computer simulation show that models are able to maintain the human comfort feeling adaptively. This research suggests an adaptive, fault-tolerant, and energy-saving control models for building spaces, using simple algorithms based on physiological adaption.