The invention provides a robot urban navigation system based on the combination of visual SLAM and a network map. The system performs the further local obstacle avoidance path optimization of an urbanroad navigation path with the help of the environment object space occupying condition, obtained through the SLAM processing, of a region where a wheel-type robot is located on the basis of an urbanroad navigation path that is determined through global path planning and along which the wheel-type robot in the network map arrives at a target place, thereby overcoming multiple technical limits anddefects of the navigation control of the wheel-type robot, only employing the global path planning, through the combination of the visual SLAM and the network map. The system can provide a finer positioning and path navigation for the wheel-type robot, and better guarantees that the wheel-type robot passes along the urban road navigation path after the local obstacle avoidance path optimization,and then better guarantees the path planning and navigation accuracy and passing effectiveness of the wheel-type robot.