The invention discloses an artificial intelligence automobile unmanned driving system. The system comprises a vehicle big data system, a road and traffic big data system, an environment perception system, a navigation system and a driving and execution control system. When an automobile travels normally on a highway, objects surrounding the automobile are recognized mainly depending on images obtained by multiple cameras, and spatial coordinates of the objects are measured to provide surrounding object information for the navigation system; at least one long-range millimeter wave radar is adopted to detect a remote target, and especially under the condition of poor visibility, for example, at night or on foggy weather, long-range early-warning information is provided for the navigation system; and when the automobile is parked, multiple short-range ultrasonic wave radars are utilized to detect short-range barriers surrounding the automobile, and sufficient geometric information is provided for an automatic parking system. Object recognition mainly depends on a deep learning system, and in cooperation with a fuzzy cognition method, object recognition speed is increased, the operation amount is reduced, the requirement on hardware is lowered, and unmanned driving response speed is increased.