In some cases crystal structure of a material, which determines all the physical properties of the material, is not known. In such circumstances, computational approaches play a crucial role. In order to predict the crystal structure of a material we developed a method called as CrystAl Structure Prediction via Simulated Annealing (CASPESA). CASPESA requires a unit cell and predefined bond constraints to start the crystal prediction. All these constraints were determined with the help of experimental structures or density functional theory calculations. Up to now, this method have been extensively employed to find the crystal structures of metal ammines and metal borohydrides which are promising materials for hydrogen storage. In these cases, the objective function in CASPESA was metal-hydrogen or chlorine-hydrogen distances. We have observed that if these distances were maximized, then the total energy of the crystal starts to lower. Of course, to apply CASPESA to an arbitrary system, first of all the following question “which interaction is is the most crucial one to obtain a lower energy structure” must be carefully addressed. After finding the most suitable objective function and several necessary bond length constraints, CASPESA is ready to run. In our group, we are still improving efficiency and robustness of CASPESA approach.