In the development of the artificial intelligent system for processing spatial data, several side problems appeared. These problems are not necessarily directly related to the main problem, but also needed to be solved in order to either evaluate or implement the algorithm. The first developed aspect is an algorithm for assessing the similarity of a grid with a given reference grid, based on how well the grid resembles the unknown underlying distribution. This approach is used in the algorithm to select parameters, but is also used to evaluate the results of different algorithms. The second development concerns a way of dealing with robustness errors when processing spatial data. The limited representation of real numbers in a computer system causes rounding errors, which—if this happens for numbers that represent coordinates—may result in robustness errors concerning the geometry calculations. The proposed method aims to find errors and circumvents them by lowering the dimension of calculated geometries. This algorithm is used throughout the implementation, as all spatial operations make use of it.