Abstract
In this study, we used robot which was the most acceptable among ´voice´, ´vision´ and ´robot´ as a driver-agent. We verified whether the driving behavior improvement by driving the driving simulator together with the agent. The agent's support was 3 kinds of "driving-support", "feedback" and "driving support and feedback". Driving-support was real-time support by robot's voice. Feedback looked back on unsafe driving behaviors with movie after driving. Driving-support and feedback was combined use of both. As a result, driving behavior was improvement by using an agent. Driving behavior was the most improved condition was "driving-support and feedback". And, the condition with higher subjective evaluation was "driving-support and feedback".
