Robotic Task
Robotic task research focuses on enabling robots to autonomously perform complex actions, often guided by high-level instructions (e.g., natural language commands). Current efforts concentrate on integrating large language models (LLMs) and vision-language models (VLMs) with task and motion planners, leveraging their semantic reasoning capabilities to generate and refine action sequences, often within a hierarchical planning framework. This research is crucial for advancing robot autonomy in unstructured environments and diverse applications, from manufacturing and warehousing to domestic assistance and search and rescue, by improving task success rates and reducing reliance on pre-programmed behaviors.
Papers
HiCRISP: An LLM-based Hierarchical Closed-Loop Robotic Intelligent Self-Correction Planner
Chenlin Ming, Jiacheng Lin, Pangkit Fong, Han Wang, Xiaoming Duan, Jianping He
Representation Abstractions as Incentives for Reinforcement Learning Agents: A Robotic Grasping Case Study
Panagiotis Petropoulakis, Ludwig Gräf, Josip Josifovski, Mohammadhossein Malmir, Alois Knoll