The Nao goal vision needs to be tested and improved. There is a base application but it has not been run in Hannover. The causes of problems need to be identified. It also needs to be checked, if the acquired data can be used as additional input for the localization. Also, if both poles of the goal are seen, a position estimate can be made directly based on this observation.

The next step is to build an goal according to the current rule book (at least in Aachen, not sure about Graz) to test the vision and then gradually improve it. For now the exact camera position is probably not required, as it has already proven in the case of the ball that the current estimate of a standing robot is good enough in the current situation. Then basic goal detection is to be tested and improved (and possibly tuned to be fast enough). Finally a basic triangulation can be added for the case where both poles are visible. This might not be necessary if this data is taken as input for the localization.