The paper presents an overview of a new originally developed quasi-optimal control algorithms for reproduction of earthquakes by shaking tables. As known, shaking tables are widely used for testing of structures, models or structural details subjected to earthquakes or other types of dynamic loadings. In order to obtain dynamic response of a tested model close to that of a real structure, the shaking table should reproduce real dynamic excitations with high accuracy. Moreover, the shaking table control algorithm should consider real restrains in the platform displacements. Changing the ground acceleration record by time scaling, like it is common in existing algorithms, yields undesired effect on the tested models dynamic behaviour. The originally developed approach, described in this paper, is based on optimal signals perturbation and filtering technique. It includes quasi-optimal spline approximation and further optimal smoothening of the dynamic loading signal, decreasing negative effects. Selecting the smoothening parameters is an adaptive procedure, performed by comparing the spectrum characteristics of the original and reproduced signals. The effectiveness of the proposed method is demonstrated by obtaining responses of multi-story structural models subjected to various seismic excitations. It is shown that the proposed algorithm is a very effective method for reproduction of real earthquakes by using shaking tables with limited stroke.