Helping People Through Space and Time: Assistance as a Perspective on Human-Robot Interaction

Benjamin A. Newman, Reuben M. Aronson, Kris Kitani, Henny AdmoniFrontiers Robotics AI 2021 · DOI · OpenReview

Abstract

The field of assistive robotics is currently fragmented, with researchers often working in domain-specific silos such as rehabilitation, education, or manufacturing. This paper addresses this lack of cohesion by proposing a unifying theoretical framework. Rather than defining assistance by specific tasks, the authors frame it as a perspective where a robot maintains autonomy over its actions while remaining subordinate to a user’s goals. Through a systematic review of over 200 papers, the study introduces three key design axes—People, Space, and Time—to provide a common language for comparing strategies and understanding design trade-offs across the diverse landscape of human-robot interaction (HRI).

Introduction

Assistive robotics research is often highly specialized. A team developing a robotic feeding arm may never consult with researchers building autonomous wheelchairs, despite both facing similar challenges in user intent recognition and safety. This fragmentation prevents the translation of breakthroughs across domains. To bridge this gap, this paper abstracts "assistance" away from specific applications, offering a general theory that allows roboticists to share solutions to universal HRI challenges.

Methodology

The authors developed their taxonomy through a systematic literature review. They initially hand-selected 74 papers from the Human-Robot Interaction (HRI) conference (2016–2020) to identify core themes and keywords. This was followed by a broader query of the Web of Science, resulting in approximately 1,500 papers. After rigorous filtering for relevance and impact, over 200 recent assistive robotics papers were classified according to the proposed framework to validate its applicability.

A New Perspective on Assistance

The core of the framework is the definition of an assistive system as one where the robot has autonomy over its actions but is subordinate to the user's goals. In this model, the human defines the objective, and the robot determines the optimal execution path to support that objective.

The People Axis: Defining Roles

In multi-person environments, a robot must distinguish between different roles to prioritize its actions effectively.

Figure 1: The People axis categorizes individuals as interactants, primary users, or additional targets of assistance.
  • Targets of Assistance: These are individuals whose goals the robot is programmed to prioritize (e.g., a patient in a clinical setting). Managing conflicting goals among multiple targets remains a significant technical and ethical challenge.
  • Interactants: These are bystanders or secondary participants. The robot must interact with them safely or navigate around them, but their goals are not the robot's primary focus (e.g., other patrons in a restaurant where a robot is assisting a specific diner).

The Space Axis: Where the Robot Acts

Assistance can be provided by acting on different dimensions of the state space:

  • Environment: The robot manipulates physical objects, such as fetching items or cleaning.
  • Human Body: The robot provides physical support or guidance, such as an exoskeleton or a guide robot for the visually impaired.
  • Human Brain: The robot assists by managing cognitive load or altering mental states, such as using social cues to guide attention or projecting visual aids onto a workspace.

The Time Axis: When the Robot Acts

The timing of assistance influences both user agency and system complexity:

  • Reactive: The robot waits for explicit user commands, maximizing user agency but increasing cognitive load.
  • Proactive: The robot anticipates user needs based on internal models, offering smoother interaction at the risk of being intrusive if predictions are incorrect.
  • Simultaneous: The robot and human act in concert, such as in shared-control tasks like collaborative lifting.

Conclusion

The proposed People-Space-Time taxonomy provides a robust structure for categorizing disparate HRI research. By highlighting that a proactive robotic drawer and an anticipatory robotic arm face identical design trade-offs regarding trust and agency, this framework encourages the development of general principles for assistive robotics that transcend specific use cases. Future work will focus on refining these axes to discover further universal challenges in the field.

Summary updated

Abstract

As assistive robotics has expanded to many task domains, comparing assistive strategies among the varieties of research becomes increasingly difficult. To begin to unify the disparate domains into a more general theory of assistance, we present a definition of assistance, a survey of existing work, and three key design axes that occur in many domains and benefit from the examination of assistance as a whole. We first define an assistance perspective that focuses on understanding a robot that is in control of its actions but subordinate to a user’s goals. Next, we use this perspective to explore design axes that arise from the problem of assistance more generally and explore how these axes have comparable trade-offs across many domains. We investigate how the assistive robot handles other people in the interaction, how the robot design can operate in a variety of action spaces to enact similar goals, and how assistive robots can vary the timing of their actions relative to the user’s behavior. While these axes are by no means comprehensive, we propose them as useful tools for unifying assistance research across domains and as examples of how taking a broader perspective on assistance enables more cross-domain theorizing about assistance.
BibTeX
@article{newman2021helping,
  author = {Benjamin A. Newman and Reuben M. Aronson and Kris Kitani and Henny Admoni},
  title = {{Helping People Through Space and Time: Assistance as a Perspective on Human-Robot Interaction}},
  journal = {Frontiers Robotics AI},
  year = {2021},
  doi = {10.3389/frobt.2021.720319},
}
Figure