A summary of the discussed issues, conclusions and answers to (some) questions that we did not have time to make during the live session can be found in this pdf.

The video of the workshop can be watched here!

Program

July 13, 2020 via Zoom. 7:00 - 9:00 Pacific Time (San Francisco), 16:00 - 18:00 CEST (Paris), 23:00 - 01:00 JST (Tokyo)

• Opening remarks/using the online system

  • 7:00 - 7:10 Pacific Time (San Francisco), 16:00 - 16:10 CEST (Paris), 23:00 - 23:10 JST (Tokyo)

• Session 1. Contributed papers

  • 7:10 - 7:40 Pacific Time (San Francisco), 16:10 - 16:40 CEST (Paris), 23:10 - 23:40 JST (Tokyo)

  • Each paper: 4 min pre-recorded teaser + 3 min questions

    • Paper 1. Perceiving Signs for Navigation Guidance in Spaces Designed for Humans [talk][pdf],

      • • Claire Liang, Cheng Perng Phoo, Laasya Renganathan, Yingying Yu, Bharath Hariharan , and Hadas Kress-Gazit

    • Paper 2. Learning with Dynamic Feedback [talk][pdf]

      • • Hang Yu and Elaine Short

    • Paper 3. Guided 360-Degree Visual Perception for Mobile Telepresence Robots [talk][pdf]

      • • Kishan Chandan, Xiaohan Zhang, Jack Albertson, Xiaoyang Zhang, Yao Liu, Shiqi Zhang

    • Paper 4. Distributed Reinforcement Learning of Targeted Grasping with Active Vision for Mobile Manipulators [talk][pdf]

      • • Yasuhiro Fujita, Kota Uenishi, Avinash Ummadisingu, Prabhat Nagarajan, Shimpei Masuda, Mario Ynocente Castro

• Session 2. Discussions with the panel

  • 7:40 - 9:00 Pacific Time (San Francisco), 16:40 - 18:00 CEST (Paris), 23:40 - 01:00 JST (Tokyo)

  • Selected questions will be made based on

    • Previously collected from Google form.

    • Made in real time (moderated). Three ways to make questions

      • Write your question on Pheedloop session of the workshop

      • Write your question on the Zoom chat

      • Raise your hand on Zoom and make the question by yourself

• Session 3. Free time

  • 9:00 - 10:00 Pacific Time (San Francisco), 18:00 - 19:00 CEST (Paris), 01:00 - 02:00 JST (Tokyo)

  • This is an optional part where people are supposed to mingle and discuss topics freely.

Despite increasing commercial demand, progress in learning methods for robotics, and the decrease in hardware costs, service robots are still far from helping us. Implicitly or explicitly, the assumptions and the way experiments are conducted in laboratories lead to research that is extremely hard to transfer to real use. As a result, scientific research becomes locked within the university environment while companies must revisit and re-validate academic work. More often than not, they develop their own methods to bring robots closer to the market. Our goal is to bring together experts in academia and industry to uncover issues that so far have bounded research to the laboratory environment and discuss the design of methods with potential towards effective, real-world deployment of service robots.

We envision the achievement of two goals: From the academic perspective, to identify research practices that align with real-world use, increasing long-term impact and practical value. From the industrial perspective, to help research scientists outside academia identify academic work with potential towards technological readiness. Researchers from both industry and academia will be encouraged to provide their insights to decrease the technical transfer gap and promote collaboration. We encourage the submission of both new research as well as short retrospectives that discuss practical limitations of past papers that the authors wish to document for future readers.

Topics of Interest

• [Naive users]: Human robot collaboration or interaction with “naive users.” In-the-wild studies or methods. What assumptions should we make of end users, and how realistic are they for real-world use?

• [Safety in physical human-robot interaction]: How can we guarantee safe interaction in the real world? How can hardware design or low cost robotics support safe interaction?

• [Learning from Demonstration and Programming by Demonstration]: How do we enable real users (who don’t know how to program) to create and modify robot behavior? Trade-offs on the simplicity vs flexibility of user interfaces

• [Retrospectives on practicality]: What limitations did you encounter in your previous research that would be useful to know for those attempting to put the work into practice?

• [Deep learning for service robotics]: How do we obtain sufficient data---or reduce the amount of data required---to apply deep learning based approaches to realistic problems? How can we close the sim2real gap and leverage simulations effectively?

• [Perception for service robots]: Realistic sensing at home, in hospitals, etc..

• [Industry research insights]:

  • Pressing issues facing real service robots

  • In-house research that academia does not provide

  • Successful and unsuccessful research cases

Panelists and (optional) supporting material

• Leila Takayama (University of California, Santa Cruz)

• Satoru Satake (ATR)

  • How to Build a Social Robot in the Real Field? Easy or Difficult?

• Pieter Abbeel (UC Berkeley, Covariant)

  • Towards AI for the Physical World

• Carlos Celemin (TU Delft)

  • Interactive Imitation Learning

• Heni Ben Amor (Arizona State University)

• Serena Ivaldi (Inria)

  • Human-robot collaboration technologies applied in real-use cases: some lessons learned

• Tetsuya Ogata (Waseda University)

  • Deep Predictive Learning for Robot Motion Generation

• Surya Singh (Intuitive Surgical)

  • * Check Intuitive's research areas/grants here!

Contributed papers

• • Perceiving Signs for Navigation Guidance in Spaces Designed for Humans [talk][pdf],

    • • Claire Liang, Cheng Perng Phoo, Laasya Renganathan, Yingying Yu, Bharath Hariharan , and Hadas Kress-Gazit

  • Learning with Dynamic Feedback [talk][pdf]

    • • Hang Yu and Elaine Short

  • Guided 360-Degree Visual Perception for Mobile Telepresence Robots [talk][pdf]

    • • Kishan Chandan, Xiaohan Zhang, Jack Albertson, Xiaoyang Zhang, Yao Liu, Shiqi Zhang

  • Distributed Reinforcement Learning of Targeted Grasping with Active Vision for Mobile Manipulators [talk][pdf]

    • • Yasuhiro Fujita, Kota Uenishi, Avinash Ummadisingu, Prabhat Nagarajan, Shimpei Masuda, Mario Ynocente Castro

Call for Papers

Important Dates (revised)

• Submission deadline for papers: June 1, 2020

• Notification of acceptance: June 15 17, 2020

• Camera-ready version: June 29, 2020

• Workshop: July 13, 2020

We invite several types of contributions:

• Full length paper: 8 pages max (excluding citations)

• Position paper: 8 pages max (excluding citations)

• Short paper: 4 pages max (excluding citations)

• Retrospective: 4 pages max (excluding citations)

Paper format: Full RSS paper format.

Submissions are not double blind

Submission website: https://easychair.org/conferences/?conf=arwgap2020

Organizers

Guilherme Maeda (Preferred Networks, Inc.) <gjmaeda(at)preferred.jp>

Nick Walker (University of Washington) <nswalker(at)cs.washington.edu>

Petar Kormushev (Imperial College London)

Maru Cabrera (University of Washington) <mecu(at)cs.washington.edu>