In this study, a three-axle vehicle model established with ADAMS/Car is first correlated with field test data from quasi-static tilt table and highly dynamic NATO double lane change maneuver tests, respectively. It is then applied to predict the vehicle static rollover threshold (SRT) and dynamic rollover threshold (DRT). With the optimization approach proposed in this study it is possible to efficiently tune the anti-roll bar stiffness at each axle, to either maximize SRT or DRT, or balance both. The sensitivity results derived from the optimization iteration process can be applied to effectively size the three anti-roll bars that balance the static and dynamic roll stability performances. The proposed method can be potentially applied to include other parameters to address the roll stability issues and beyond.