Walking with robots: Video analysis of human-robot interactions in transit spaces

Paper Title

Walking with robots: Video analysis of human-robot interactions in transit spaces

Publication Info

• Topic area: Human-robot interaction in public transit spaces
• Keywords: Human-robot interaction, social navigation, transit spaces, cleaning robots, video analysis, ethnomethodology, conversation analysis, robot design, public spaces, socially-aware movement

Background and Problem

• Problem / challenge: Current robots deployed in public spaces, such as cleaning robots, lack the interactional competence to navigate complex social environments smoothly. They fail to understand human movement, social group dynamics, and the purpose of specific spaces.
• Significance: As robots become more common in public spaces, their inability to integrate into social environments causes disruptions, inefficiencies, and additional labor for human staff. Addressing these issues is critical for their effective deployment.
• Motivation and related work: Prior research has focused on technical aspects of robot navigation, such as collision avoidance and proxemics, but has largely overlooked the social and interactional dimensions of movement. Ethnographic studies have highlighted the challenges of robot integration in public spaces, but a gap remains in designing robots that can understand and adapt to social contexts.

Solution

• Proposed approach: The study uses video analysis of cleaning robots in a commercial airport to identify design challenges and propose the concept of "socially-aware movement" for robots.
• Novelty:
  1. Demonstrates how robots disrupt social order in transit spaces due to a lack of social understanding.
  2. Proposes three "strong concepts" for designing socially competent robots: sequentiality of movement, membership categorization, and scenic intelligibility.
  3. Highlights the need for a shift from technical to socio-technical design in robot navigation.
• Procedure and key techniques:
  • Conducted six hours of video-ethnographic fieldwork in a European airport, observing two cleaning robots (Nilfisk Liberty SC60 and Ecobot 40).
  • Analyzed 40 interaction highlights using ethnomethodology and conversation analysis to identify patterns and challenges in robot-human interactions.
  • Developed design recommendations based on observed social and spatial dynamics.

Results

• Concrete findings:
  • Robots often halt unpredictably, causing confusion and requiring humans to adjust their movements ("halting problem").
  • Robots fail to recognize social groups (e.g., queues) and disrupt their organization.
  • Robots lack understanding of the functional purposes of spaces (e.g., toilets, waiting areas), leading to disruptive behavior.
• Advantage over baselines: The study provides a richer understanding of the social and sequential nature of movement, which is absent in traditional technical approaches to robot navigation.
• Experiments / evaluation:
  • Video recordings of robots cleaning in three airport areas.
  • Analysis of 22 selected clips illustrating interactional challenges.
  • Observations supplemented by interviews with cleaning staff and airport personnel.
• Limitations and future work:
  • Small dataset limited to two robots in one airport.
  • Findings are specific to transit spaces and may not generalize to other public contexts.
  • Future work should explore long-term studies, diverse settings, and technical implementation of the proposed concepts.

Summary

This study investigates the challenges of deploying cleaning robots in a public airport, revealing their inability to navigate social environments effectively. Using video analysis, it identifies key issues such as the "halting problem," failure to recognize social groups, and lack of understanding of space functionality. The authors propose three "strong concepts"—sequentiality of movement, membership categorization, and scenic intelligibility—to guide the design of socially-aware robots. These findings highlight the need for a socio-technical approach to robot navigation, emphasizing human-centric design for seamless integration into public spaces.