In order to improve steering and roll stability for a heavy vehicle, a novel MIMO (multi-input-multi-output) controller is proposed based on an AMFC theory. A vehicle model considering the lateral motion, yaw motion and roll motion is developed. The establishment of reference models, controlled models and iteration algorithm of the feed-back matrix are presented. The Newton iteration method is used to correct the feed-back matrix in the real time, and the ideal control signal is calculated by the integral feed-forward matrix. The state of controlled model can approach reference model to achieve an adaptive online update for the steering control parameters under the transient conditions. The proposed AMFC have a good improvement in reducing side-slip angle, yaw rate and roll angle. The simulation and HIL experiments demonstrate the effectiveness and reliability of the proposed controller in improving the yaw stability and thus preventing rollover when necessary. [ABSTRACT FROM AUTHOR]