In this paper we propose a behavior representation system for robotic wheelchairs in which behavior information regarding the robotic wheelchair is provided to passengers to improve their sense of comfort; physiological indices are also used to measure passenger comfort. Many boarding-type robots (i.e., autonomous robots that carry passengers, such as autonomous vehicles, personal mobility robots, and robotic wheelchairs) have already been developed already. Most studies regarding this type of robot focus on safety control or efficient path planning but fail to consider passenger comfort. This paper describes how factors leading to passenger discomfort can be avoided by sharing behavior information. This study evaluates the validity of information sharing by estimating a stress level measured using heart rate and skin conductance as physiological indices. To investigate the relationship between different discomforting events and passengers' physiology indices, an experiment is conducted in which each passenger's physiology indices are measured while the autonomous robotic wheelchair is moving. For long-lasting stress events, the significant difference (p < 0:01) in heart rate is shown by behavior representing of a robotic wheelchair with these indices. For strongly discomforting events, the significant difference (p < 0:01) in skin response is also shown. These results show that the proposed system is effective for maintaining passenger comfort in robotic wheelchairs.