Challenges in Applying Academic Research to Real-World Robotics

Full-day workshop at ICRA 22; May 23, 2022.

Abstract

There are tremendous commercial opportunities in applying robotic capabilities to solve real-world problems; ranging from agriculture and medicine to manufacturing and autonomous vehicles. Exploiting these opportunities, however, still requires significant hardware and software innovation coupled with cutting-edge, academic research. Directly applying results published in the academic literature can be challenging for many commercial applications due to the misalignment of requirements between the settings. For example, consider safety, robustness, and reproducibility.

The industrial engineer must ensure that their product is robust and meets governmental safety guidelines from bodies such as the NHTSA. If these safety ratings are not demonstrated, it does not matter how effective or novel the solution is. It can not be sold and, therefore, is worthless. Similarly, if a commercial product does not repeatedly and reliably perform its function under, oftentimes, non-ideal conditions, it will not be a scalable, profitable product and will fail when presented to the customer. The academic researcher is working outside of such product safety constraints. They, of course, maintain safety in their lab for their students and associates but do not need to consider product safety requirements as part of their research. A similar argument may be made around robustness and reproducibility. Obtaining a novel result is a primary requirement for many publications. Therefore, robustness and reproducibility are deemphasized in favor of continual advancement and gives little academic insight into the engineering effort required to reproduce or adapt approaches to a commercial application.


This workshop looks at the competing drivers in academia and industry settings that influence modern robotics research and highlights the challenges industry roboticists face when attempting to apply published, academic research. A balance of invited talks and facilitated discussions will encourage participation from individuals across the robotics subfields. Invited speakers will share the approaches and challenges they’ve when attempting to integrate academic research to a commercial product. A facilitated panel discussion will delve further into shared experience and explore ideas for better transferring academic research advancements to industry settings.

Call for Abstracts

See page

Confirmed Speakers

Schedule [full playlist of recorded talks]

09:00 - Introduction [link to recording]

09:15 - Speaker: Jonathan Tompson [link to recording]

09:45 - Speaker: Marco Hutter and Christian Gehring [link to recording]

10:15 - Break

10:45 - Speaker: Nicolai Anton Lynnerup [link to recording]

11:15 - Speaker: Nikhil Mishra [link to recording]

11:45 - Mini-break/panel setup

12:00 - Morning panel discussion [link to recording]

12:30 - Lunch

13:45 - Speaker: Melonee Wise [link to recording]

14:15 - Speaker: Bharath Sankaran [link to recording]

14:45 - Lightning talks

15:15 - Break

15:45 - Speaker: George Konidaris [link to recording]

16:15 - Speaker: Damion Shelton [link to recording]

16:35 - Afternoon panel discussion [link to recording]

17:00 - Close

Venue

The workshop was held at the Pennsylvania Convention Center in Philadelphia, PA, as well as, virtually over Zoom on 23 May, 2022. The workshop was recorded, and you can find a playlist of all the talks and panel discussions here.

Organizers

Connor Schenck (Third Wave Automation)

Arunkumar Byravan (DeepMind)

James Davidson (Talos Robotics)

Scott Kuindersma (Boston Dynamics)