The aim of this paper is to elaborate a reliable rollover risk evaluator in order to supply reliable information to the control system or to alert the driver and advise him on a speed limit. This is done by calculating the probability of rollover risk of heavy vehicle. The evaluation is based on rollover risk indicator, namely load transfer ratio (LTR) corresponding to each axle. Before doing this, sensitivity analysis of the driver-vehicle-infrastructure system is given in order to find the most influencing parameters on the risk, such as the center height of gravity and the cornering stiffness. On another hand, some of these parameters are time varying and not directly measured. So, two methods are used to identify all parameters of heavy vehicle: an Unscented Kalman Filter (UKF) is developed to estimate all dynamic states and to identify some parameters. And maximum entropy principle (MEP) is used to probabilistic modeling of the other parameters. Finally, the probability laws obtained by MEP and UKF are used to determine the probability of the risk.