부산대학교 도서관

In this paper, we propose RBT-HCI, a reliable behavior tree (BT) planning method with human-computer interaction, aiming at generating an interpretable and human-acceptable BT. Compared with other BT generation methods, RBT-HCI can reliably plan a BT based on the knowledge base. When an available BT cannot be planned automatically, instead of terminating or relaxing the rules, RBT-HCI provides a new idea, which is to make decisions through human-computer interaction, thereby enhancing the reliability and robustness of the method. The effectiveness of RBT-HCI is verified by an example of robot grasping objects, showing that a reliable and robust planning result can be obtained through knowledge-based automatic planning and human-computer interaction.