Long Horizon Task Planning

Long-horizon task planning aims to enable robots to execute complex, multi-step tasks requiring extended temporal reasoning and adaptation. Current research heavily utilizes large language models (LLMs) coupled with various scene representations (e.g., 3D scene graphs) to generate and refine plans, often incorporating iterative feedback loops and hierarchical planning structures. This work is significant because it pushes the boundaries of robotic autonomy, enabling robots to handle more complex and unpredictable real-world scenarios, with applications ranging from household assistance to industrial automation.