IEEE ICRA 2022 Workshop
Reinforcement Learning for Contact-Rich Manipulation
May 27, Philadelphia (PA), USA
The workshop will be held in person with support for remote presentation and participation.
Important Details
Submission deadline:
April 22, 2022 (AoE)April 30, 2022 (AoE)Paper (extended abstract) format: IEEE RAS format
Recommended length: maximum 4 pages excluding references
Submission link: OpenReview
Notification of acceptance:
April 30, 2022May 6, 2022Workshop date: May 27, 2022
Location: Room 118 C
Contact-rich manipulation tasks are an important and challenging category of tasks in robotic automation for a variety of applications. Physical contact is prevalent in object insertion and assembly. For tasks involving complex contact dynamics and friction, it is difficult to model related physical effects, and hence, traditional control methods often result in inaccurate or brittle controllers. Lately, reinforcement learning (RL) has been demonstrated to be a promising approach to learning robot control policies in such environments. However, RL still faces many challenges in contact-rich environments, such as in sample efficiency, sim2real transfer, safety and stability, and reward shaping.
The objective of the workshop is to focus on addressing the challenges specifically involved in contact-rich environments and gather researchers in this area to share ideas and state-of-the-art solutions.
Call for Contributions
We invite submissions from a broad range of topics, including but not limited to:
Imitation learning
Learning from CAD
Multimodal representations
Reward learning
Safety and stability guarantees
Sample efficiency
Sim2real transfer
Tactile representation
Task sequence learning
Accepted contributions will be presented as lightning talks and posters. The two best contributions will be presented as spotlight talks.
Invited Speakers
Stanford University
Osaka University
New York University
UC Berkeley
Accepted Papers
Pathologies and Challenges of Using Differentiable Simulators in Policy Optimization for Contact-Rich Manipulation. H.J. Terry Suh, Max Simchowitz, Kaiqing Zhang, Tao Pang, Russ Tedrake. (oral paper)
SAGCI-System: Towards Sample-Efficient, Generalizable, Compositional, and Incremental Robot Learning. Jun Lv, Qiaojun Yu, Lin Shao, Wenhai Liu, Wenqiang Xu, Cewu Lu. (oral paper)
RRL: Resnet as representation for Reinforcement Learning. Rutav Shah, Vikash Kumar. (oral paper)
Self-Supervised Learning of Multi-Object Keypoints for Robotic Manipulation. Jan Ole von Hartz, Eugenio Chisari, Tim Welschehold, Abhinav Valada.
Learning to Grasp the Ungraspable with Emergent Extrinsic Dexterity. Wenxuan Zhou, David Held.
Efficient Object Manipulation Planning with Monte Carlo Tree Search. Huaijiang Zhu, Ludovic Righetti.
Tactile Sensing and its Role in Learning and Deploying Robotic Grasping Controllers. Alexander Koenig, Zixi Liu, Lucas Janson, Robert Howe.
Learning active tactile perception through belief-space control. Jean-François Tremblay, Johanna Hansen, David Meger, Francois Robert Hogan, Gregory Dudek.
Learning Slip with a Patterned Capacitive Tactile Sensor. Yuri Gloumakov, Tae Myung Huh.
Learning Goal-Oriented Non-Prehensile Pushing in Cluttered Scenes. Nils Dengler, David Großklaus, Maren Bennewitz. (remote presentation) (poster zoom link)
Integrating Force-based Manipulation Primitives with Deep Learning-based Visual Servoing for Robotic Assembly. Yee Sien Lee, Nghia Vuong, Nicholas Adrian, Quang Cuong Pham. (remote presentation) (poster zoom link)
Learning Dense Reward with Temporal Variant Self-Supervision. Yuning Wu, Jieliang Luo, Hui Li. (remote presentation) (poster zoom link)
Synthesizing and Simulating Volumetric Meshes from Vision-based Tactile Imprints. Xinghao Zhu, Siddarth Jain, Masayoshi Tomizuka, Jeroen Vanbaar. (remote presentation) (poster zoom link)
Schedule
09:00 - 09:15 Welcome and introduction
09:15 - 09:45 Invited talk: Jeannette Bohg - Fusing Vision and Touch for Contact-Rich Manipulation
09:45 - 10:15 Invited talk: Kensuke Harada - Application of Reinforcement Learning for Force Controlled Tasks (remote)
10:15 - 10:40 Coffee break
10:40 - 11:10 Invited talk: Aude Billard - TBA
11:10 - 11:20 Spotlight talk: Pathologies and Challenges of Using Differentiable Simulators in Policy Optimization for Contact-Rich Manipulation
11:20 - 11:50 Invited talk: Shabhaz Khader - Control Stability in Learning Contact-Rich Manipulation Skills (remote)
11:50 - 12:20 Invited talk: Ludovic Righetti - A few ideas to improve efficiency and robustness of complex contact interactions
12:20 - 13:45 Lunch break
13:45 - 14:15 Invited talk: Sergey Levine - Data-Driven Robotic Reinforcement Learning
14:15 - 14:25 Spotlight talk: SAGCI-System: Towards Sample-Efficient, Generalizable, Compositional, and Incremental Robot Learning
14:25 - 14:35 Spotlight talk: RRL: Resnet as representation for Reinforcement Learning
14:35 - 15:05 Lightning talks
15:05 - 16:00 Coffee break & Posters (in person and remote)
16:00 - 16:15 Final remarks
Program Committee
Christian Pek
KTH
Cristian C. Beltran-Hernandez
Osaka University
Ioanna Mitsioni
KTH
Krishnan Srinivasan
Stanford University
Nghia Vuong
Nanyang Technological University
Shaoxiong Wang
MIT
Zheng Wu
UC Berkeley
Organizers