Workshop
Generalizable Policy Learning in the Physical World
Young Min Kim 路 Sergey Levine 路 Ming Lin 路 Tongzhou Mu 路 Ashvin Nair 路 Hao Su
Fri 29 Apr, 8 a.m. PDT
While the study of generalization has played an essential role in many application domains of machine learning (e.g., image recognition and natural language processing), it did not receive the same amount of attention in common frameworks of policy learning (e.g., reinforcement learning and imitation learning) at the early stage for reasons such as policy optimization is difficult and benchmark datasets are not quite ready yet. Generalization is particularly important when learning policies to interact with the physical world. The spectrum of such policies is broad: the policies can be high-level, such as action plans that concern temporal dependencies and causalities of environment states; or low-level, such as object manipulation skills to transform objects that are rigid, articulated, soft, or even fluid.In the physical world, an embodied agent can face a number of changing factors such as \textbf{physical parameters, action spaces, tasks, visual appearances of the scenes, geometry and topology of the objects}, etc. And many important real-world tasks involving generalizable policy learning, e.g., visual navigation, object manipulation, and autonomous driving. Therefore, learning generalizable policies is crucial to developing intelligent embodied agents in the real world. Though important, the field is very much under-explored in a systematic way.Learning generalizable policies in the physical world requires deep synergistic efforts across fields of vision, learning, and robotics, and poses many interesting research problems. This workshop is designed to foster progress in generalizable policy learning, in particular, with a focus on the tasks in the physical world, such as visual navigation, object manipulation, and autonomous driving. We envision that the workshop will bring together interdisciplinary researchers from machine learning, computer vision, and robotics to discuss the current and future research on this topic.
Schedule
Fri 8:00 a.m. - 8:10 a.m.
|
Introduction and Opening Remarks
(
Introduction
)
>
|
Hao Su 馃敆 |
Fri 8:10 a.m. - 8:35 a.m.
|
Invited Talk (Danica Kragic): Learning for contact rich tasks
(
Invited Talk
)
>
SlidesLive Video |
Danica Kragic 馃敆 |
Fri 8:35 a.m. - 8:40 a.m.
|
Q&A for Invited Talk (Danica Kragic)
(
Q&A
)
>
|
Danica Kragic 馃敆 |
Fri 8:40 a.m. - 9:05 a.m.
|
Invited Talk (Peter Stone): Grounded Simulation Learning for Sim2Real
(
Invited Talk
)
>
SlidesLive Video |
Peter Stone 馃敆 |
Fri 9:05 a.m. - 9:10 a.m.
|
Q&A for Invited Talk (Peter Stone)
(
Q&A
)
>
|
Peter Stone 馃敆 |
Fri 9:10 a.m. - 9:20 a.m.
|
Break
|
馃敆 |
Fri 9:20 a.m. - 10:15 a.m.
|
Poster Session 1 ( Poster Session ) > link | 馃敆 |
Fri 10:15 a.m. - 11:15 a.m.
|
Panel Discussion
(
Panel Discussion
)
>
|
Young Min Kim 路 Peter Stone 路 Nadia Figueroa 路 Hao Su 路 Mrinal Kalakrishnan 路 Xiaolong Wang 路 Deepak Pathak 路 Ming Lin 路 Danfei Xu 馃敆 |
Fri 11:15 a.m. - 11:23 a.m.
|
ManiSkill Challenge Winner Presentation (Zhutian Yang & Aidan Curtis)
(
Contributed Talk
)
>
SlidesLive Video |
Zhutian Yang 馃敆 |
Fri 11:23 a.m. - 11:31 a.m.
|
ManiSkill Challenge Winner Presentation (Fattonny)
(
Contributed Talk
)
>
SlidesLive Video |
Kun Wu 馃敆 |
Fri 11:31 a.m. - 1:00 p.m.
|
Lunch Break
|
馃敆 |
Fri 1:00 p.m. - 1:10 p.m.
|
Contributed Talk (Sim-to-Lab-to-Real: Safe RL with Shielding and Generalization Guarantees)
(
Contributed Talk
)
>
SlidesLive Video |
Kai-Chieh Hsu 馃敆 |
Fri 1:10 p.m. - 1:35 p.m.
|
Invited Talk (Shuran Song): Iterative Residual Policy for Generalizable Dynamic Manipulation of Deformable Objects
(
Invited Talk
)
>
SlidesLive Video |
Shuran Song 馃敆 |
Fri 1:35 p.m. - 1:40 p.m.
|
Q&A for Invited Talk (Shuran Song)
(
Q&A
)
>
|
Shuran Song 馃敆 |
Fri 1:40 p.m. - 2:05 p.m.
|
Invited Talk (Nadia Figueroa): Towards Safe and Efficient Learning and Control for Physical Human Robot Interaction
(
Invited Talk
)
>
SlidesLive Video |
Nadia Figueroa 馃敆 |
Fri 2:05 p.m. - 2:10 p.m.
|
Q&A for Invited Talk (Nadia Figueroa)
(
Q&A
)
>
|
Nadia Figueroa 馃敆 |
Fri 2:10 p.m. - 2:18 p.m.
|
ManiSkill Challenge Winner Presentation (EPIC Lab)
(
Contributed Talk
)
>
SlidesLive Video |
Weikang Wan 馃敆 |
Fri 2:18 p.m. - 2:30 p.m.
|
Break
|
馃敆 |
Fri 2:30 p.m. - 2:40 p.m.
|
Contributed Talk (Know Thyself: Transferable Visual Control Policies Through Robot-Awareness)
(
Contributed Talk
)
>
SlidesLive Video |
Edward Hu 馃敆 |
Fri 2:40 p.m. - 3:05 p.m.
|
Invited Talk (Mrinal Kalakrishnan): Robot Learning & Generalization in the Real World
(
Invited Talk
)
>
SlidesLive Video |
Mrinal Kalakrishnan 馃敆 |
Fri 3:05 p.m. - 3:10 p.m.
|
Q&A for Invited Talk (Mrinal Kalakrishnan)
(
Q&A
)
>
|
Mrinal Kalakrishnan 馃敆 |
Fri 3:10 p.m. - 3:35 p.m.
|
Invited Talk (Xiaolong Wang): Generalizing Dexterous Manipulation by Learning from Humans
(
Invited Talk
)
>
SlidesLive Video |
Xiaolong Wang 馃敆 |
Fri 3:35 p.m. - 3:40 p.m.
|
Q&A for Invited Talk (Xiaolong Wang)
(
Q&A
)
>
|
Xiaolong Wang 馃敆 |
Fri 3:40 p.m. - 3:48 p.m.
|
ManiSkill Challenge Winner Presentation (Silver-Bullet-3D)
(
Contributed Talk
)
>
SlidesLive Video |
Yingwei Pan 馃敆 |
Fri 3:48 p.m. - 3:50 p.m.
|
Break
|
馃敆 |
Fri 3:50 p.m. - 4:45 p.m.
|
Poster Session 2 ( Poster Session ) > link | 馃敆 |
Fri 4:45 p.m. - 5:30 p.m.
|
ManiSkill Challenge Award Ceremony
(
Challenge Award Ceremony
)
>
|
13 presentersHao Su 路 Weikang Wan 路 Hao Shen 路 He Wang 路 Yingwei Pan 路 Zhutian Yang 路 Fabian Dubois 路 Tom Sonoda 路 Kun Wu 路 Kangqi Ma 路 Liu Kun 路 Jilei Hou 路 Tongzhou Mu |
Fri 5:30 p.m. - 6:30 p.m.
|
Closing Remarks
(
Closing Remarks
)
>
|
馃敆 |
-
|
PAnDR: Fast Adaptation to New Environments from Offline Experiences via Decoupling Policy and Environment Representations ( Poster ) > link | Sang Tong 路 Hongyao Tang 路 Yi Ma 路 Jianye HAO 路 YAN ZHENG 路 Zhaopeng Meng 路 Boyan Li 路 Zhen Wang 馃敆 |
-
|
Imitation Learning for Generalizable Self-driving Policy with Sim-to-real Transfer ( Poster ) > link | Zolt谩n L艖rincz 路 M谩rton Szemenyei 路 Robert Moni 馃敆 |
-
|
FlexiBiT: Flexible Inference in Sequential Decision Problems via Bidirectional Transformers ( Poster ) > link |
11 presentersMicah Carroll 路 Jessy Lin 路 Orr Paradise 路 Raluca Georgescu 路 Mingfei Sun 路 David Bignell 路 Stephanie Milani 路 Katja Hofmann 路 Matthew Hausknecht 路 Anca Dragan 路 Sam Devlin |
-
|
Learning Category-Level Generalizable Object Manipulation Policy via Generative Adversarial Self-Imitation Learning from Demonstrations ( Poster ) > link | Hao Shen 路 Weikang Wan 路 He Wang 馃敆 |
-
|
A Study of Off-Policy Learning in Environments with Procedural Content Generation ( Poster ) > link | Andrew Ehrenberg 路 Robert Kirk 路 Minqi Jiang 路 Edward Grefenstette 路 Tim Rocktaeschel 馃敆 |
-
|
Planning to Practice: Efficient Online Fine-Tuning by Composing Goals in Latent Space ( Poster ) > link | Kuan Fang 路 Patrick Yin 路 Ashvin Nair 路 Sergey Levine 馃敆 |
-
|
Learning Transferable Policies By Inferring Agent Morphology ( Poster ) > link | Brandon Trabucco 路 mariano Phielipp 路 Glen Berseth 馃敆 |
-
|
Using Deep Learning to Bootstrap Abstractions for Robot Planning ( Poster ) > link | Naman Shah 路 Siddharth Srivastava 馃敆 |
-
|
Don't Freeze Your Embedding: Lessons from Policy Finetuning in Environment Transfer ( Poster ) > link | Victoria Dean 路 Daniel Toyama 路 Doina Precup 路 Victoria Dean 馃敆 |
-
|
Safer Autonomous Driving in a Stochastic, Partially-Observable Environment by Hierarchical Contingency Planning ( Poster ) > link | Ugo Lecerf 路 Christelle Yemdji-Tchassi 路 Pietro Michiardi 馃敆 |
-
|
Separating the World and Ego Models for Self-Driving
(
Poster
)
>
link
SlidesLive Video |
Vlad Sobal 路 Alfredo Canziani 路 Nicolas Carion 路 Kyunghyun Cho 路 Yann LeCun 馃敆 |
-
|
Multi-objective evolution for Generalizable Policy Gradient Algorithms ( Poster ) > link | Juan Jose Garau-Luis 路 Yingjie Miao 路 John Co-Reyes 路 Aaron Parisi 路 Jie Tan 路 Esteban Real 路 Aleksandra Faust 馃敆 |
-
|
ShiftNorm: On Data Efficiency in Reinforcement Learning with Shift Normalization
(
Poster
)
>
link
SlidesLive Video |
Sicong Liu 路 Xi Zhang 路 Yushuo Li 路 Yifan Zhang 路 Jian Cheng 馃敆 |
-
|
Improving performance on the ManiSkill Challenge via Super-convergence and Multi-Task Learning ( Poster ) > link | Fabian Dubois 路 Eric Platon 路 Tom Sonoda 馃敆 |
-
|
Multi-task Reinforcement Learning with Task Representation Method ( Poster ) > link | Myungsik Cho 路 Whiyoung Jung 路 Youngchul Sung 馃敆 |
-
|
Deep Sequenced Linear Dynamical Systems for Manipulation Policy Learning ( Poster ) > link | Mohammad Nomaan Qureshi 路 Ben Eisner 路 David Held 馃敆 |
-
|
Learning Robust Task Context with Hypothetical Analogy-Making ( Poster ) > link | Shinyoung Joo 路 Sang Wan Lee 馃敆 |
-
|
Silver-Bullet-3D at ManiSkill 2021: Learning-from-Demonstrations and Heuristic Rule-based Methods for Object Manipulation ( Poster ) > link | Yingwei Pan 路 Yehao Li 路 Yiheng Zhang 路 Qi Cai 路 Fuchen Long 路 Zhaofan Qiu 路 Ting Yao 路 Tao Mei 馃敆 |
-
|
Zero-Shot Reward Specification via Grounded Natural Language ( Poster ) > link | Parsa Mahmoudieh 路 Deepak Pathak 路 trevor darrell 馃敆 |
-
|
Reinforcement Learning for Location-Aware Warehouse Scheduling ( Poster ) > link | Stelios Stavroulakis 路 Biswa Sengupta 馃敆 |
-
|
A Probabilistic Perspective on Reinforcement Learning via Supervised Learning ( Poster ) > link | Alexandre Piche 路 Rafael Pardinas 路 David Vazquez 路 Chris J Pal 馃敆 |
-
|
Prompts and Pre-Trained Language Models for Offline Reinforcement Learning ( Poster ) > link | Denis Tarasov 路 Vladislav Kurenkov 路 Sergey Kolesnikov 馃敆 |
-
|
Compositional Multi-Object Reinforcement Learning with Linear Relation Networks ( Poster ) > link | Davide Mambelli 路 Frederik Tr盲uble 路 Stefan Bauer 路 Bernhard Schoelkopf 路 Francesco Locatello 馃敆 |
-
|
Density Estimation For Conservative Q-Learning ( Poster ) > link | Paul Daoudi 路 Ludovic Dos Santos 路 Merwan Barlier 路 Aladin Virmaux 馃敆 |
-
|
Control of Two-way Coupled Fluid Systems with Differentiable Solvers
(
Poster
)
>
link
SlidesLive Video |
Brener Ramos 路 Felix Trost 路 Nils Thuerey 馃敆 |
-
|
One-Shot Imitation with Skill Chaining using a Goal-Conditioned Policy in Long-Horizon Control ( Poster ) > link | Hayato Watahiki 路 Yoshimasa Tsuruoka 馃敆 |
-
|
Versatile Offline Imitation Learning via State-Occupancy Matching ( Poster ) > link | Yecheng Jason Ma 路 Andrew Shen 路 Dinesh Jayaraman 路 Osbert Bastani 馃敆 |
-
|
Let鈥檚 Handle It: Generalizable Manipulation of Articulated Objects ( Poster ) > link | Zhutian Yang 路 Aidan Curtis 馃敆 |
-
|
Revisiting Model-based Value Expansion ( Poster ) > link | Daniel Palenicek 路 Michael Lutter 路 Jan Peters 馃敆 |
-
|
An Empirical Study and Analysis of Learning Generalizable Manipulation Skill in the SAPIEN Simulator ( Poster ) > link | Liu Kun 路 Huiyuan Fu 路 Zheng Zhang 路 huanpu yin 馃敆 |
-
|
Don't Change the Algorithm, Change the Data: Exploratory Data for Offline Reinforcement Learning ( Poster ) > link | Denis Yarats 路 David Brandfonbrener 路 Hao Liu 路 Michael Laskin 路 Pieter Abbeel 路 Alessandro Lazaric 路 Lerrel Pinto 馃敆 |
-
|
Learning Generalizable Dexterous Manipulation from Human Grasp Affordance ( Poster ) > link | Yueh-Hua Wu 路 Jiashun Wang 路 Xiaolong Wang 馃敆 |
-
|
Continuous Control on Time ( Poster ) > link | Tianwei Ni 路 Eric Jang 路 Tianwei Ni 馃敆 |
-
|
A Minimalist Ensemble Method for Generalizable Offline Deep Reinforcement Learning ( Poster ) > link | Kun Wu 路 Yinuo Zhao 路 Zhiyuan Xu 路 Zhen Zhao 路 Pei Ren 路 Zhengping Che 路 Chi Liu 路 Feifei Feng 路 Jian Tang 馃敆 |
-
|
Know Thyself: Transferable Visual Control Policies Through Robot-Awareness ( Poster ) > link | Edward Hu 路 Kun Huang 路 Oleh Rybkin 路 Dinesh Jayaraman 馃敆 |
-
|
Sim-to-Lab-to-Real: Safe RL with Shielding and Generalization Guarantees ( Poster ) > link | Kai-Chieh Hsu 路 Allen Z. Ren 路 Duy Nguyen 路 Anirudha Majumdar 路 Jaime Fern谩ndez Fisac 馃敆 |