The IEEE/ACM Conference on Human Robot Interaction (HRI 2023) is 2 weeks away. As we wrote before, HRI is an exciting conference showcasing all the latest research and technology progress in the effectiveness of humans working with robots. It is also one of the conferences we are happy to extensively cover for only our paid subscribers. An immense amount of funding and effort is invested into HRI, so if you can afford a paid subscription, please join us.

This post discusses two pieces of work which are already featured by their respective authors on Twitter. The full schedule of HRI 2023 is available here. The compete details of workshops, which usually feature latest cutting edge research which couldn’t make it in time for the main conference, is available here.

Select pieces of work on Twitter

“No, to the Right” — Online Language Corrections for Robotic Manipulation via Shared Autonomy” from reseachers at Stanford University

Siddharth Karamcheti @siddkaramcheti

Want to build robots that adapt to language corrections in real-time? Introducing "No, to the Right – Online Language Corrections for Manipulation via Shared Autonomy" (arxiv.org/abs/2301.02555) w/ @YuchenCui1, Raj, Nidhya, @percyliang & @DorsaSadigh at #HRI2023 - 🧵👇 (1/N).

4:21 PM ∙ Jan 9, 2023

---

91Likes26Retweets

The full paper is available here. I found this work very interesting; the main idea is that the performance of a robot in accomplishing a task such as placing a book on the shelf can be remarkably improved in real time with the help of verbal prompts and suggestions from a human. The robot them learns from how it corrected its mistakes, and completes the task correctly in subsequent attempts. The authors demonstrate that with their suggested technique called Language-Informed Latent Actions with Corrections (LILAC), complex tasks can be learned by a robot in 10-20 demonstrations
instead of the thousands to tens of thousands of demonstrations required by fully autonomous imitation or reinforcement learning approaches.

“Crossing Reality: Comparing Physical and Virtual Robot Deixis” from researchers at Colarodo School of Mines

Zhao Han @hanzhao

Non-verbal cues like pointing are proven for robots to communicate. But many robots don't have arms to point! We investigated #AugmentedReality arm. Are they better than physical arms? & virtual arms pointing at physical objects? 👉Our #HRI2023 paper: zhaohanphd.com/HRI23AR

zhaohanphd.comCrossing Reality: Comparing Physical and Virtual Robot Deixis - Zhao Han, Ph.D.Augmented Reality (AR) technologies present an exciting new medium for human-robot interactions, enabling new opportunities for both implicit and explicit

5:12 PM ∙ Feb 21, 2023

---

6Likes2Retweets

This work considers how virtual arms created using Augmented Reality (AR) can be used to help robots and drones which lack arms and need to use them. The researchers use a TurtleBot and Microsoft HoloLens 2 Mixed Reality Set in their study and show that an AR virtual arm can achieve the same level of performance and also be as warm and likable as a physical arm attached to the robot. This work seems impactful because it shows how AR based solutions can be used to extend the capability of robots which do not have a head or arms to interactively communicate with humans. A short presentation of this work is available here.

Conclusion

We will keep documenting more pieces of work discussed at HRI 2023. We will also be attending the HRI 2023 conference virtually, and hope to bring in more interesting insights to you. Thank you once again for paying for this newsletter and allowing us to offer you our services.