Requirement Analysis of Lidar Sensor Setups for Self-Localization in Automated Valet Parking

In the use case of automated valet parking (AVP), localization must be accomplished based on environmental sensors due to the lack of satellite visibility. Previous research has presented suitable methods, which, on the other hand, generally apply a 360° lidar setup. In order to achieve a series introduction of AVP, it is necessary from an economic point of view to achieve self-localization with scaled-down sensor setups that correspond to the current state of series production. Addressing this challenge, this work investigates to what extent a 360° lidar reference sensor setup can be reduced in order to achieve sufficiently robust self-localization in parking garage environments. The investigations are carried out in a real environment in two parking garages in conjunction with a test vehicle equipped with a 360° lidar reference setup. On the basis of a definition of different sensor arrangements, a parameter variation is carried out. As evaluation basis for the requirement analysis, a self-localization based on a modified iterative closest point (ICP) algorithm introduced in the scope of the work is applied.