Agile Robotics: From Perception to Dynamic Action
@ ICRA2024

Robots today often navigate or manipulate slowly and intermittently, relying on bursts of computation for planning. This is in contrast to humans who routinely use agile and dynamic motions such as vaulting power cords over chairs when vacuuming, flinging garments to flatten before folding, lofting sheets across beds, jumping across obstacles, and running for fun or to reach a destination on time. Dynamic motions can speed task completion and manipulate objects out of reach, but they require careful integration of grasp planning, motion planning, and time-parameterization, lifting quasi-static assumptions, and strong safety considerations.

The objectives of the workshop are to understand how we can get agile robots that operate faster and reliably. We wish for the workshop scope to include the full spectrum from system design to algorithms and techniques to applications of agile robots. By doing so, we aim to broaden the appeal of the workshop. We hope that the proposed workshop will lead to a deeper understanding of agile robots and facilitate collaborations between participants, speakers, and organizers to advance the field.

The workshop will be held in hybrid mode

Date: May 13, 2024

Links:

Content

Topics

Agility, characterized by swift, precise, and adaptive movement in complex environments, is crucial for robots to fully exploit their capabilities. An agile robot can leap over obstacles, manipulate objects out of reach, navigate through tight spaces, and quickly adapt to unexpected changes. Historically, research in agile robotics has often been segmented. Whether it's a focus on specific platforms like arms, legs, or drones; a preference for classical control versus machine learning methods; or a distinct emphasis on either perception or low-level control, there's been a tendency to dive deep into individual areas. However, achieving true robotic agility demands an integrated approach that bridges these traditional divides. This workshop seeks to foster discussions that view agility as a comprehensive objective, necessitating collaboration across domains and platforms. The scope of the workshop thus focuses primarily on (but is not limited to) the following topics:

  • System Design
    • Hardware and mechanical considerations specific to agile robots.
    • Unified agility through the integration of traditionally distinct robotic domains, such as whole-body mobile manipulation.
    • Synchronization of grasp planning, motion planning, and time parameterization.
  • Techniques and Challenges
    • Integrating recent machine learning techniques, such as large language models, foundational models, and transformers, into agile robotics.
    • Role of perception in enhancing robot agility.
    • Stability, safety, and evaluation of robots operating at high speeds.
  • Applications
    • Showcasing agility in real-world scenarios: fast bin-picking, quick assembly, speedy delivery robots, accurate object tossing and catching, and more.

Schedule


Time Zone: GMT +09

Time Activity
09:00 - 09:15 Workshop opening
09:15 - 09:35 Invited Talk #1
09:35 - 09:55 Invited Talk #2
09:55 - 10:15 Invited Talk #3
10:15 - 10:45 Spotlight talks #1
  • Speakers contents
10:45 - 11:30 Coffee break + Posters session #1
  • Posters
11:30 – 11:50 Invited Talk #4
11:50 – 12:10 Invited Talk #5
12:10 – 13:15 Lunch
13:15 – 13:35 Invited Talk #6
13:35 – 13:55 Invited Talk #7
13:55 – 14:15 Invited Talk #8
14:15 – 14:45 Spotlight talks #2
  • Speakers contents
14:45 – 15:30 Coffee break + Posters session #2
  • Posters
15:30 – 15:50 Invited Talk #9
15:50 – 16:10 Invited Talk #10
16:10 – 16:30 Invited Talk #11
16:30 – 17:00 Panel discussion
17:00 – 17:15 Closing remarks

Call for papers

Paper Submission Deadline: 29 March 2024, 23:59pm Pacific Time

Notification of Acceptance: 12 April 2024

We encourage extended abstract submission up to 3+n pages, with n pages (no page-limit) for the bibliography or supplementary materials, in the IEEE conference style. Any type of contribution, either already published or work in progress, is welcome to our workshop.

Accepted papers will be presented in a poster session and selected papers as spotlight talks. All submitted contributions will go through a single-blind review process. The contributed papers will be made available on the workshop’s website. However, this does not constitute an archival publication and no formal workshop proceedings will be made available, meaning contributors are free to publish their work in archival journals or conferences.

To submit papers:

https://openreview.net/group?id=IEEE.org/2024/ICRA/Workshop/Agile_Robotics

Invited Speakers (alphabetical order)

Pulkit Agrawal

Pulkit Agrawal


Assistant Professor
Massachusetts Institute of Technology

Talk title: Dynamic, Dexterous and Purposeful Robotic Control


Aude Billard

Aude Billard


Professor
École Polytechnique Fédérale de Lausanne

Talk title: TBD


Jemin Hwangbo

Jemin Hwangbo


Professor
Korea Advanced Institute of Science and Technology

Talk title: Dynamic Legged Locomotion and Manipulation


Masatoshi Ishikawa

Masatoshi Ishikawa


Professor
Tokyo University of Science

Talk title: High-speed Robots Based on High-speed Visual Feedback


Aaron Johnson

Aaron Johnson


Associate Professor
Carnegie Mellon University

Talk title: Learning How to Drive Over Extreme Terrain


Dimitrios Kanoulas

Dimitrios Kanoulas


Associate Professor
University College London

Talk title: Cognitive Real-World Loco-manipulation


Quan Nguyen

Quan Nguyen


Assistant Professor
University of Southern California

Talk title: Hierarchical Optimization-base Control for Agile Legged Robots


Deepak Pathak

Deepak Pathak


Assistant Professor
Carnegie Mellon University

Talk title: Rethinking Reinforcement Learning for Extremely Agile and Dynamic Robotic Control


Shuran Song

Shuran Song


Assistant Professor
Stanford University

Talk title: Dynamics Manipulation for Deformable Objects


Hao Su

Hao Su


Associate Professor
University of California, San Diego

Talk title: Learning Generalizable Manipulation Skills from 3D Input


Xiaolong Wang

Xiaolong Wang


Assistant Professor
University of California, San Diego

Talk title: 3D Scene Understanding for Locomotion Control


Organizers (alphabetical order)

Contact

If you have any questions please contact Jeff Ichnowski at the email: jeffi AT cmu DOT edu