An emotion-based software robot for automobiles, in which a driver's emotion and behavior caused by such emotion are anticipated when each input such as a driver's states, commands, and behaviors, automobile situations, automobile environmental situations, etc., is recognized based on results learned with respect to a change in emotion of each individual driver offline, as well as each piece of vehicle information, is assigned a priority, so that services provided by a telematics system, etc., can be selectively implemented to conform to a driver's mood.