The feasibility of a real-time, rule-based guidance system for an automobile on a limited-access highway is studied. The inputs to the system are parameters describing the traffic situation, road geometry, vehicle state, and a driver selected goal. The output of the system consists of guidance parameters that command control logic governing the throttle, steering angle, and brakes. During past research, a dynamic expert system accomplished this guidance task under the assumption that all inputs were available deterministically, i.e., without error. An investigation is made of the types and effects of uncertainties encountered in a more realistic environment, and a stochastic methodology is proposed to extend the expert system to handle information that is known only in a probabilistic way.