PROBLEM TO BE SOLVED: To improve the communication function with a driver, by deciding the action of an anthropomorphic agent based on a learned result through learning means and a status judged by a status judging means, and making an appeared agent perform the decided action. SOLUTION: A navigation processing part 10 of a general control part 1, which controls the entire communication reaches a course to a set destination point for guiding in voice or with an image display. An agent processing part 11 learns car's status and past address by a driver, etc., for appropriate conversation and control. An image processing part 13 processes such input image as anthropomorphic agent image, map image, etc., from the navigation processing part 10 and the agent processing part 11 through an I/F part 12. A sound voice control part 14 control such sound voice output as agent sound voice and course guiding sound voice and inputted sound voice, while a status information processing part 15 processes a detection data of various status regarding a car and a driver.