Learning for Caregiving Robots

Overview

YouTube Recording: The recording for the Learning for Caregiving Robots workshop is now live on YouTube! A huge thank you again to all of the speakers and attendees.

Slack (for Zoom link, discussion, and Q/A): https://join.slack.com/t/learningforca-9j78633/shared_invite/zt-qlb8sgph-MAl26xEFPSCoQLaxuPaq2Q

Robotic caregivers could increase the independence of people with disabilities, improve quality of life, and help address growing societal needs, such as shortages of healthcare workers, aging populations that require care, and high healthcare costs. Using robotic technology to assist care-recipients in their daily lives may entail physically interacting with them, adapting to their preferences, and perceiving the environment for intelligent and safe assistance. These tasks can benefit from using data-driven machine learning techniques. Through this workshop, we explore how learning can enable robotic systems towards achieving consistently efficient and safe assistance across activities of daily living.

Learning for caregiving in real homes depends on sensing and acting using real-world systems as well as metrics for success and generalization. Through a series of three correspondingly themed sessions, this workshop will provide a forum for discussion on the state-of-the-art, current challenges, and future of learning paradigms in robotic caregiving. By bringing together renowned scientists and young researchers from the machine learning and assistive robotics communities, this workshop aims to inspire cross-discipline collaborations towards solutions that can beneficially impact human society.

Important Dates

Paper Submission Deadline: April 23, 2021. April 30, 2021
Notification of Acceptance: April 30, 2021. May 7, 2021
Camera-Ready Deadline: May 21, 2021
Workshop Date: May 31, 2021

Call for Contributions

Congratulations to our Best Presentation Award winner!

Affordance-Aware Handovers with Human Arm Mobility Constraints
Paola Ardón, Maria E. Cabrera, Èric Pairet, Ronald P. A. Petrick, Subramanian Ramamoorthy, Katrin S. Lohan, and Maya Cakmak

We welcome submissions across the domains of machine learning, robotic caregiving, human-robot interaction, haptics, biomechanics, perception, simulation, and human assistance. As we aim to encourage meaningful discussion across these fields, work that is unpublished, recently published or under review can be accepted for presentation. Best Workshop Presentation will be given to an outstanding contribution, with a $1000 award, generously sponsored by Hello Robot.

We solicit contributions in the form of:

• Extended abstracts (min 2 pages, max 4 pages) in IEEE paper format.

All PDF submissions (and video demo links) should be sent to: lcr.icra2021@gmail.com

• Papers will be presented at the workshop with accompanying spotlight talks (and workshop posters pending in-person participation).
• Authors are welcome to submit a video demo with their contribution.
• Accepted paper contributions will appear on the workshop website. Authors retain all the intellectual properties of their contributions.
• Award certificate and a $1000 award for the Best Workshop Presentation will be given to an outstanding contribution.

Topics of Interest

• Robotic caregiving / physical human assistance
• Learning
• Human-robot collaboration
• Physical human-robot interaction
• Personalization and adaptation
• Haptic sensing
• Multimodal perception
• Active perception
• Human pose sensing/estimation
• Deep learning for safe human-robot interaction
• Physics simulation and benchmarking
• Generalization of learned models to real-world assistive scenarios

Invited Speakers

	Prof. Pieter Abbeel, University of California, Berkeley, US. Title: Vision-based Reinforcement Learning for Caregiving Robots Abstract: Learning from visual observations is a fundamental yet challenging problem in reinforcement learning. Although algorithmic advancements combined with convolutional neural networks have proved to be a recipe for success, it’s been widely accepted that learning from pixels is not as efficient as learning from direct access to underlying state. However, for practical caregiving robots, it is important they can directly learn from visual input. In this talk I will describe our recent work that (almost entirely) bridges the gap in sample complexity between learning from pixels and from state, as empirically validated on the DeepMind Control Suite and Atari games. In fact, I will present two new approaches establishing this new state of the art: Reinforcement Learning with Augmented Data (RAD) and Contrastive Unsupervised Representations for Reinforcement Learning (CURL). At the core of both are data augmentation through random crops. Our approaches outperform prior pixel-based methods, both model-based and model-free, on complex tasks in the DeepMind Control Suite and Atari Games showing 1.9x and 1.6x performance gains at the 100K environment and interaction steps benchmarks respectively. I will also present preliminary results on the task of assistive dressing.
	Prof. Henny Admoni, Carnegie Mellon University, US. Title: The Role of Human Behavior in Caregiving Robots Abstract: Physically and socially assistive robots can improve health and quality of life. To be effective, these robots must be able to understand their human partners, for example by recognizing their goals, predicting their next actions, and identifying when they need help. Much of the information about these internal mental states is revealed nonverbally, through eye gaze, gestures, and other behaviors that provide implicit signals. Therefore, assistive robots must understand and use nonverbal communication. This requires a multidisciplinary approach that involves robotics, psychology, machine learning, and computer vision. In this talk, I will describe my work on robots that assist humans on complex tasks, such as eating a meal or engaging in a conversation. Throughout the talk, I will describe how techniques and knowledge from cognitive science help us develop robot algorithms that lead to more effective interactions between people and their robot partners.
	Prof. Brenna Argall, Northwestern University, US. Title: Machine Learning for Assistive Partial Autonomy Abstract: As need increases, access decreases. It is a paradox that as human motor impairments become more severe, and increasing assistance needs are paired with decreasing motor abilities, the very machines created to provide this assistance become less and less accessible to operate with independence. My lab addresses this paradox by incorporating robotics autonomy and intelligence into physically-assistive machines: leveraging robotics autonomy, to advance human autonomy. Achieving the correct allocation of control between the human and the autonomy is essential, and critical for adoption. The allocation must be responsive to individual abilities and preferences, that moreover can be changing over time, and robust to human-machine information flow that is filtered and masked by motor impairment and control interface. As we see time and again in our work and within the field: customization and adaptation are key, and so the opportunities for machine learning are clear.
	Prof. Michelle Johnson, University Of Pennsylvania, US. Title: From Human-Human Interactions in Task-Oriented Stroke Therapy to Robot-Human Interactions Abstract: Autonomous and semi-autonomous robots have been increasingly proposed as social and therapy assistants for neurorehabilitation after traumatic and non-traumatic brain injury. Unfortunately, current robot-patient interactions do not accurately model therapist-patient interactions in task-oriented stroke rehabilitation and therefore may fall short of being clinically effective. We coded and analyzed 8 videos each showing an occupational therapist interacting with a stroke patient in a common activities of daily living setting. We propose that a model of the interaction between a patient and a therapist can be overlaid on a stimulus-response paradigm where the therapist and the patient take on a set of acting states or roles and are motivated to move from one role to another when certain physical or verbal stimuli or cues are sensed and received. I will discuss how the observed roles and cues can be mapped to current and future examples of robot-patient interactions and ethical implications if such a robot was realized.
	Prof. Dongheui Lee, Technical University of Munich, Germany. Title: Human centric Assistive Robotics from collaboration to postural control Abstract: In this talk I will introduce some of our recent developments on human centric approaches for assistive technology. Understanding human movement is a key to design human centered assistance solutions. For that, we either model human movements using a template model or learn joint human-robot tasks by following learning from demonstration paradigm. The applications range from teleoperation-based assistance to joint object transportation and to human postural control.
	Prof. C. Karen Liu, Stanford University, US. Title: Simulating Human Movements for Assistive Robotics Abstract: Unlike conventional robots, robots that physically assist humans need to embrace physical contacts and learn to utilize them for enabling human’s activities of daily living. An immediate concern in developing such an autonomous and powered robotic device is the safety of human users during the early development phase when the control policies are still largely suboptimal. Learning from physically simulated humans and environments presents a promising alternative which enables robots to safely make and learn from mistakes without putting real people at risk. However, deploying such policies to interact with people in the real world adds additional complexity to the already challenging sim-to-real transfer problem. In this talk, I will present our current progress on solving the problem of sim-to-real transfer with humans in the environment, actively interacting with the robots through physical contacts. We tackle the problem from two fronts: developing more relevant human models to facilitate robot learning and developing human-aware robot perception and control policies. To contextualize our research effort, we develop a mobile manipulator to put clothes on people with physical impairments, enabling them to carry out day-to-day tasks and maintain independence.
	Dr. Jeremy Marvel, National Institute of Standards and Technology, US. Title: Toward Consistent Metrics in Caregiver HRI Abstract: The field of human-robot interaction lacks consistency in the selection and utility of metrics for evaluating system, interface, and teaming performance. Rather than using a consistent and broadly applicable base set of test methods and measurements, researchers often leverage customized methodologies chosen to highlight specific capabilities, and downplay limitations. In this talk, Dr. Jeremy Marvel from the National Institute of Standards and Technology will discuss the steps necessary to move toward test methods and metrics that can be widely and consistently applied as tools for evaluating, benchmarking, verifying, and validating human-robot systems in caregiver applications.
	Prof. Laurel Riek, UC San Diego, US. Title: How community health perspectives can help us build better robots Abstract: Our society does not adequately support people with disabilities or their families, which leads to large rates of burnout, and ever-increasing health and financial inequities. Robots are not going to fix these major societal problems; however, our work explores how we might be able to design technology that lessens the burden of systemic ableism. We do this by centering and including people with disabilities, their caregivers, and their communities as we envision, design, develop, and deploy future robotic systems. I will discuss several of our recent projects in community health contexts, where we are creating interactive and adaptive systems that aim to extend the reach of cognitive neurorehabilitative therapies, provide respite to overburdened caregivers, and explore how technology might serve as a means for mediating positive interactions during hardship.
	Prof. Siddhartha Srinivasa, University of Washington, US. Title: Robot Assisted Feeding – A Problem Every Roboticist Should be Working on Abstract: This talk is not about a solution but about a problem. Feeding is an essential activity of daily living. It’s also often an engaging and pleasurable social group interaction. However, a significant portion of people with upper extremity disabilities cannot feed by themselves. Although robotics might just be part of a holistic solution, I believe it is imperative for us roboticists to scrutinize this problem and contribute to its solution. I’ll talk about some of the technical challenges of robot assisted feeding, specifically around bite acquisition and around bite transfer. Together, these problems encompass robotics, machine learning, and human-robot interaction (and possibly more fields).

Organizers

	Zackory Erickson, PhD Student in Robotics, Georgia Institute of Technology, US.
	Dr. Tapomayukh Bhattacharjee, Post-doc, University of Washington (UW); Assistant Professor, Cornell University (Summer 2021), US.
	Prof. Carme Torras, Professor of Research, Institut de Robòtica i Informàtica Industrial, CSIC-UPC, Spain.
	Prof. Yiannis Demiris, Professor in Human-Centred Robotics, Imperial College London, UK.
	Prof. Charlie Kemp, Associate Professor in the Department of Biomedical Engineering, Georgia Institute of Technology, US. Conflict of Interest Statement: Prof. Kemp owns equity in and works part time for Hello Robot Inc., a company that commercialized robotic assistance technologies developed in his lab.
	Prof. Tomohiro Shibata, Professor of Graduate School of Life Science and Systems Engineering, Kyushu Institute of Technology, Japan.

Workshop Program

Time (ET) (GMT-04)
10:00 – 10:15	Welcome – Workshop opening
10:15 – 11:15	Opening Session – Learning for Robotic Caregivers: Perspectives from 3 Continents (US, Europe, Asia) – Prof. Tomohiro Shibata, Kyushu Institute of Technology, Japan. – Prof. Yiannis Demiris, Imperial College London, UK. – Dr. Tapomayukh Bhattacharjee, University of Washington, US.
(Session) Sensing and Acting: intelligent sensing, systems and control for robot caregivers
11:15 – 11:45	Prof. Henny Admoni, Carnegie Mellon University, US.
11:45 – 12:15	Prof. Laurel Riek, UC San Diego, US.
12:15 – 12:45	Prof. Brenna Argall, Northwestern University, US.
12:45 – 13:00	Break
13:00 – 13:15	Contributor Presentation Antonio Andriella, Learning In Situ Personalised Robot Assistance From Therapist’s Demonstrations and Knowledge
13:15 – 13:30	Contributor Presentation Paola Ardón, Affordance-Aware Handovers with Human Arm Mobility Constraints
13:30 – 14:00	Panel Discussion – The role of learning and autonomy in robotic caregiving Panelists Henny Admoni Laurel Riek Dongheui Lee Yiannis Demiris
14:00 – 15:15	Lunch break
(Session) Learning: roles, applications, and generalization for robot caregivers
15:15 – 15:45	Prof. Dongheui Lee, Technical University of Munich, Germany.
15:45 – 16:15	Prof. Siddhartha Srinivasa, University of Washington, US.
16:15 – 16:45	Prof. Pieter Abbeel, University of California, Berkeley, US.
16:45 – 17:00	Presentation by Hello Robot Dr. Aaron Edsinger, Hello Robot, US.
17:00 – 17:45	SpatialChat and Break
(Session) Modeling and Benchmarking: modeling and benchmarking for robot caregivers
17:45 – 18:15	Prof. C. Karen Liu, Stanford University, US.
18:15 – 18:45	Dr. Jeremy Marvel, National Institute of Standards and Technology, US.
18:45 – 19:15	Prof. Michelle Johnson, University Of Pennsylvania, US.
19:15 – 19:45	Panel Discussion – Grand challenges in robotic caregiving Panelists Karen Liu Jeremy Marvel Michelle Johnson Aaron Edsinger
19:45 – 20:00	Awards Ceremony and Closing

Sponsor

Support and Endorsement

IEEE RAS technical committee on Robot Learning

IEEE RAS technical committee on Rehabilitation and Assistive Robotics

IEEE RAS technical committee on Human-Robot Interaction and Coordination