March 11, 2022

Human-Interactive Robot Learning (HIRL)

at HRI'22

Workshop recordings here.

With robots poised to enter our daily environments, we conjecture that they will not only need to work for people, but also learn from them. An active area of investigation in the robotics, machine learning, and human-robot interaction communities is the design of teachable robotic agents that can learn interactively from human input, including demonstrations, feedback, advice, corrections, instructions, etc. To refer to these research efforts, we use the umbrella term Human-Interactive Robot Learning (HIRL).

While algorithmic solutions for robots learning from people has been investigated in a variety of ways, HIRL, as a fairly new research area, is still lacking:

1) a formal set of definitions to classify related but distinct research problems or solutions,

2) benchmark tasks, interactions, and metrics to evaluate the performance of HIRL algorithms and interactions, and

3) clear long-term research challenges to be addressed by different communities.

The main goal of this workshop is to consolidate relevant recent works falling under the umbrella of HIRL into a coherent set of long, medium, and short-term research problems, and identify the most pressing future research goals. This workshop will also be used to develop and share diverse benchmark tasks and metrics for HIRL. These tasks would ideally span different HRI settings that are not necessarily restricted by the robotic platform.

While there has been an increased interest in the topics proposed for this workshop, and many relevant papers have been published in the last five years, research efforts falling under the HIRL umbrella are disparate across different communities. Thus, we conjecture that this is the right time to set up a workshop that will bring together researchers from these different communities with complementing research agendas to consolidate the lessons that were learned so far, and offer benchmarks to evaluate contributions on. These benchmarks, in turn, will enable researchers to have more productive collaborations and opportunities to better compare their methods and results.

As this workshop aims to draw researchers with some relevant experience, it will first consist of presentations of new or recently published works. Then, the presenters will be divided into working groups, whose participants and discussion topics will be curated according to their interests. Other people who wish to contribute to the discussions are welcome to join one of the study groups that will be formed.

Invited Speakers

Shiwali Mohan

Xerox PARC

Matthew Taylor

University of Alberta

Andrea Thomaz

University of Texas at Austin

Frank Krueger

George Mason University

Brian Scassellati

Yale University

Matthias Scheutz

Tufts University

The topics of interest for this workshop include, but are not limited to:

  • Learning from demonstration, learning by imitation, or learning from observation

  • Inverse reinforcement learning

  • Interactive reinforcement learning

  • Robot learning from human feedback (including advice, corrections, etc.)

  • Standardized task development for HIRL

  • Evaluation metrics for learners and teachers

  • Social signal processing for human teaching behaviors

  • Natural teaching interfaces

  • Teacher-learner adaptation

  • Human-guided exploration

  • Human-in-the-loop lifelong learning


Reuth Mirsky

The University of Texas at Austin and Bar Ilan University

Kim Baraka

Vrije Universiteit Amsterdam

Taylor Kessler Faulkner

The University of Texas at Austin

Justin Hart

The University of Texas at Austin

Harel Yedidsion

Applied Materials

Xuesu Xiao

X, The Moonshot Factory