Due to the different properties of the dynamical simulation of multi-body systems, it is important to introduce a classification that allows for the right selection of a simulation type according to the users needs. The proposed classification divides simulations with respect to the main characteristics, such as speed, accuracy, processing power, preparation work, resulting in the need for very different models of the rover. Using the SOLERO rover as an example, problems are shown that can occur when developing a rover model, e.g. hyper-statism. Applying simulation techniques, the rover geometry can be optimized regarding special metrics and the performance of different rovers can be compared. Experimental results of the implementation of a torque control algorithm underline the need and gain of the use of simulations in a development project.