Ensuring safety and reliability in modern vehicles requires rigorous testing of new systems across a wide range of scenarios and conditions. This fact highlights the need for effective methods and tools to evaluate performance, adaptability, and compliance with safety standards in advancing automotive technologies. MIL, SIL and HIL testing are well stablished model-based tools in the development of brake and integrated chassis control systems, and they must include the commands and reactions from the driver to fully close the system loop. Our objective is to develop a driver model with realistic behavior and that is easily adapted to different driving styles, road conditions, tracks, dynamic tests and vehicles. Part of the complexity in the development of driver models is the difference between a controller for an autonomous vehicle, and that of simulating the behavior of a human driver. The objective is no longer optimally following a track but doing it with the same control actions as a human would do it. In consequence, although the driver model structure resembles that of an autonomous vehicle, the inner algorithms are adapted to resemble a human behavior. This difference affects both the design of the algorithms and the validation of the driver model afterwards. This work presents a generic driver model for performing track and dynamic tests on a full-vehicle real-time simulation environment. It is aimed at replicating realistic (non-optimal) driver behavior, including delays and limited planning capabilities inherent to human performance. The model is based only on a few physically based parameters, which are used to adapt the driving conditions from calm to aggressive styles. Furthermore, it requires no extensive training to get realistic and stable performance, so it is easily adaptable to varying tracks and vehicles. Particularly, the driver model consists of an algorithm that computes the necessary control actions to ensure the vehicle following a given track while adhering to the imposed dynamic constraints for representing the driver. This is achieved through a three-stage process: 1) An off-line analysis both of the track, to extract curvature information, and of the vehicle to learn the longitudinal dynamics; 2) A lateral and longitudinal on-line reference generation based on track characteristics (assuming a flat track), vehicle states, and driver parameters such as look-out distance (limited forward track distance used for planning), maximum speed, and acceleration limits; 3) The controller stage, where the generated references are used to determine the appropriate control signals (steering wheel angle, throttle, and braking pedal positions). To analyze and validate the driver model, it is implemented in a full-vehicle simulation environment. The system validation is carried out by comparing simulation data with experimental data, focusing on key metrics such as speed, lateral acceleration, lateral error, and driver control signals. The presented model demonstrates a robust performance across various tracks and vehicles, maintaining a good correlation with real driver behavior obtained from experimental tests, including single lane change and track tests. It effectively adapts to different driver conditions, such as calm or aggressive driving styles, and various track scenarios while ensuring stability and accurate track following. The main contribution of this work is that, unlike existing approaches in the literature, the algorithm generates online longitudinal and lateral references for the controllers, based on the system states, the desired dynamic driver parameters and the track curvature information computed off-line. This real-time reference generation enables the system to adapt to various driving styles and conditions, without requiring extensive model training on a specific track or vehicle. The generality of the driver is assured by the utilization of physically based parameters for the longitudinal and lateral controllers. Apart from that, the inherent learning capability of the driver is obtained by a feedforward law, which is offline adjusted based on simple maneuvers very much like those naturally performed by a human driver. Regarding future development, while the current model uses the centerline as a reference for track following, future improvements will incorporate the full track width to enable a more realistic path generation. Finally, while some controller parameters require further tuning, the model demonstrates good track following and a stable trajectory, delivering strong results even on tracks with elevation changes. Further testing on more complex, longer tracks and additional drivers will help refine its performance.