Design of multiple artificial potential function and selector function for autonomous driving vehicle

This study proposed a novel method for designing artificial potential functions (APFs) for autonomous driving vehicle. In this method, a control performance index is firstly designed that reflects the control requirements. In practical use, a suitable APF should be selected from the prepared candidate APFs. In designing an APF selector function, it is necessary to include the value of performance index in the learning process to exclude dangerous choice of candidate APF caused by misclassification. It can be realized by defining the novel error function in the learning process of neural networks. The usefulness of the proposed method is verified through numerical experiments.