This research deals with the construction of a driver model for carrying out simulation on performance evaluation of obstacle avoidance on CAE. In this model construction, the risk sensed by the driver is defined as the risk potential, and the course determination and the feedback control that can pass through the obstacle are described using the potential. The experiments for obstacle avoidance are carried out, and risk potentials are constructed based on this result. Next, a course is decided so as to pass through the minimum point of these potentials, and a driver model is constructed by feed-forward control for that course and feedback control using risk potential. Finally, we confirm its characteristics by simulation.