The objective of this work is to develop a realistic driver model which helps in simulating drive related behavior of system vehicle and other vehicles in a traffic simulation environment. A driver model is said to be realistic only if it can learn and adapt to any variations in vehicle parameters and simulated road conditions. At the same time, the control action and the learning should represent human-like computation. In this paper, the proposed driver model consists of a Self-Learning Model Reference Fuzzy Longitudinal and Lateral controller. The model employs a set of fuzzy rules to realize a path-following lateral controller whereas the longitudinal control is governed by another set of fuzzy rules. The adaptive capabilities of the model are realized using supervisory fuzzy set and simple self-learning algorithm. This adaptive mechanism evaluates the current controller performance against the desired closed loop reference model. Based on the evaluation results, the scaling factors are tuned according to a set of supervisory fuzzy rules. The learning happens as long as deviation of the controller performance with respect to closed loop reference model is not within a specified tolerance. Also, it happens concurrently with the control action and so it can monitor and prevent any undesirable performance at any time. Similar to human predictive action, the learnt gain values are stored and scheduled next time when relevant vehicle settings and road conditions are recognized.