Complex Robotic Task

Complex robotic task execution is a rapidly advancing field aiming to enable robots to perform intricate, multi-step actions autonomously. Current research focuses on improving learning efficiency through techniques like curriculum learning (often guided by large language models), hierarchical reinforcement learning, and imitation learning, often incorporating model-based and model-free approaches alongside advanced vision and tactile sensing. These advancements are crucial for bridging the gap between simulated and real-world robotic performance, ultimately leading to more robust and adaptable robots for industrial automation, assistive technologies, and other applications.

Papers

November 9, 2022

Leveraging Sequentiality in Reinforcement Learning from a Single Demonstration
Alexandre Chenu, Olivier Serris, Olivier Sigaud, Nicolas Perrin-Gilbert
Reinforcement Learning Deep Reinforcement Learning Video Demonstration Goal Conditioned Policy Complex Robotic Task

April 15, 2022

Divide & Conquer Imitation Learning
Alexandre Chenu, Nicolas Perrin-Gilbert, Olivier Sigaud
Imitation Learning Manipulation Task Data Divide Complex Robotic Task

March 8, 2022

Learning Sensorimotor Primitives of Sequential Manipulation Tasks from Visual Demonstrations
Junchi Liang, Bowen Wen, Kostas Bekris, Abdeslam Boularias
Manipulation Task Video Demonstration Motion Primitive Task Demonstration Sequential Manipulation Complex Robotic Task

Complex Robotic Task

Papers

Leveraging Sequentiality in Reinforcement Learning from a Single Demonstration

Divide & Conquer Imitation Learning

Learning Sensorimotor Primitives of Sequential Manipulation Tasks from Visual Demonstrations