Virtual vehicle testing on combustion engine test benches is becoming increasingly important to enhance testing speed in the automobile industry. To be able to achieve this target, dynamical test benches must be used which allow reproducing the load conditions on the engine crankshaft as occurring in the real vehicle. Dynamical testing usually consists in tracking speed versus torque profiles, which represent the expected vehicle operation. This also allows testing conditions which can not be measured in a real vehicle, and can lead to the situation in which the dynamics limit of the test bench and of the combustion engines are trespassed. This often leads to a chaotic performance of the test bench and to unreliable results. To cope with this problem we present an online capable algorithm which ensures that the test cycle is inside the performance limits. The algorithms are based on prefiltered design, the computation of feasible sets and of the use of augmented Kalman Filters to enforce feasibility. To prove the efficiency of the approach the algorithms are presented and simulation results are shown. © 2006 IEEE.