This paper focuses on the method and implementation of a map generator for the USARSim environment, which is capable of generating indoor environments. The generator adapts to a difficulty measure, which signifies how difficult the generated map should be, when mapped by a robot. Both the method of the procedural generation process as well as the knowledge on the difficulty measure are explained, followed by the implementation of the generator. Multiple maps with various difficulties are generated and mapping runs are simulated by experienced robot operators. Then the difficulty is assessed by these operators and compared to the difficulty level of the maps. The rules of the generator turn out to be able to influence the difficulty of the maps, but due to the complexities of ’difficulty’ it is difficult to do this consistently.