This work represents an historical introduction of the Robust Geometric Computation problem (RGC) and a state of the art report of the different methods to avoid the inaccuracies that usually may lead to inconsistent decisions, causing unexpected failures for some correct input data. We define the RGC problem as the trouble of developing programs that do not crash and that provide the expected outcomes. It does not necessary implies an exact computation of the numerical values. This target could be reached in several scopes with different methods, therefore we also propose a taxonomy.