The development and preliminary testing of an algorithm that detects impending rollover is described. The algorithm is intended for use in conjunction with active chassis systems to prevent some types of rollovers. Several methods of estimating roll angle and roll rate from physical principles using available sensors are presented. Advantages and drawbacks of each method and their ranges of reliable operations are discussed. An approach using a closed-loop adaptive observer for estimating roll angle and roll rate of vehicle body with respect to the road is proposed. Performance and robustness of the estimator are evaluated through numerical simulations and vehicle testing, using various scenarios, including maneuver induced and road induced body roll. The estimates are shown to be consistently superior to those obtained from other methods. A roll index is proposed, which uses the estimated roll angle and other signals to indicate rollover danger. It is shown to have good discriminatory capability in avoiding false activations and detecting impending rollovers.