This paper presents a new method to design a fuzzy controller, called `stochastic fuzzy control', instead of using traditional fuzzy reasoning, and describes the application of this method in controlling a mobile robot driven by two independent wheels. The stochastic fuzzy controller mainly consists of two parts, the observer that estimates the initial state and the feedback controller. Using this scheme, the fuzzy control system can be realized as an optimal stochastic control which ensures the stability of the control system. Two design methods are considered: one assumes that the control object is completely known, the other assumes that the control object includes some unknown parameters. We design the reference trajectories of straight line and circular path in accordance with the practical applications for the mobile robot, in which three patterns of the speed are adopted. The effectiveness of the proposed method is demonstrated by a series of practical tests on our experimental mobile robot.