The thesis concerns the automatization of hardware optimization for grouting robots. The control policy for a given robot configuration has to train automatically to evaluate the benefit/detriment of hardware changes on the performance. Create an approximation of a grouting robot in simulation, create a range of different rooms to be grouted, and propose a reward function to train a reinforcement learning grouting policy. Define a subspace of possible hardware designs for the grouting robots. Build a system to evaluate the utility of any design from the subspace and use it in search of a better hardware configuration for the task of grouting. The thesis should also present a survey of related literature.