In this paper, a nonlinear control approach of an innovative engine cooling system for vehicles is presented. The electrically driven coolant pump and a servo-controlled bypass valve as control inputs, however, are subject to saturation due to physical limitations of the maximum pump volume flow and the limited opening section of the bypass valve. Based on a control-oriented system representation, a robust decentralized control employing sliding-mode techniques is proposed: an input-output linearizing control is designed for the engine outlet temperature, whereas an exact linearization is performed for the engine inlet temperature. At this control design, the given actuator limitations are explicitly taken into account. The controllers are implemented with small sampling time and combined with a discrete-time Extended Kalman Filter that estimates unknown heat flows within the system. In an experimental investigation, the performance of two alternative stabilizing control laws is compared: a linear stabilizing control law as reference and the robust sliding-mode approach leading to a nonlinear error dynamics. The obtained results highlight the effectiveness and the control performance of the proposed robust control strategy.