Hierarchical Task Planning for Human-Robot Collaborative Assembly

In this thesis, we address the challenges of task representation and task planning for assembly within a communication-free Human-Robot Collaboration (HRC) framework. Our primary goal is to present a novel task planning approach that does not rely on modeling the human as a predictable agent. Instead, we treat the human as an unpredictable agent and focus on optimizing the robot's actions. Building on the existing Hierarchical Task Network (HTN) planning framework, UHTP (User-aware Hierarchical Task Planning), we present an implementation based on personal interpretation, extending it to handle joint actions and incorporate a failure recovery process. The secondary goal is to develop a method for autonomously retrieving task representations from annotated videos, organizing tasks into a hierarchical structure. The proposed solutions are validated through simulation experiments and compared with others to demonstrate their robustness and effectiveness.