"Hello, I'm Delivering. Let Me Pass By": Navigating Public Pathways with Walk-along with Robots in Crowded City Streets

1. Problem Statement

The rapid proliferation of autonomous mobile robots (AMRs), specifically delivery robots, in uncontrolled urban public spaces presents a significant challenge for Human-Robot Interaction (HRI) research.

• Limitations of Current Methods: Existing HRI field studies often rely on controlled experiments, "Wizard of Oz" techniques (where a human remotely controls a robot), or short-term observations in pre-designated areas. These methods fail to capture the dynamic, unpredictable nature of commercial robots operating in real-world environments.
• The Research Gap: Researchers struggle to define the scope of observation in open environments: Who should be observed (direct interactors vs. incidental passersby)? Where should researchers position themselves? How long should observation last? Traditional methods often treat robots as passive objects or focus solely on human-centric interactions, missing the robot's agency and its role in shaping the environment.
• The Need: A methodology is required to study autonomous robots that operate beyond researcher control, navigating complex social and physical landscapes where variables (weather, crowds, obstacles) are unpredictable.

2. Methodology: Walk-along with Robots (WawR)

The authors propose Walk-along with Robots (WawR), an ethnographic methodology inspired by "mobile methods" in urban studies, sociology, and geography (e.g., the "go-along" method).

• Core Concept: Researchers physically walk alongside autonomous delivery robots during their operational hours, treating the robot as an active agent and guide rather than a passive subject.
• Theoretical Framework:
  • More-than-Human Perspective: Drawing on Anna Tsing's work, the method treats robots as "interlocutors" and active participants that shape the field, rather than just tools used by humans.
  • Mobilities Paradigm: It integrates the study of movement, tempo, and the social context of urban spaces, viewing the robot's path as a network of human and non-human actors.
• Implementation Steps:
  1. Mapping & Familiarization: Researchers map robot routes using service apps and conduct "auto-ethnography" by ordering food to experience the robot's path as a user.
  2. Embodied Participation: Researchers position themselves at robot docking stations ("pre-operation waiting areas") and follow the robot on foot.
  3. Continuous Observation: Unlike fragmented studies, WawR tracks the robot's full operational cycle (departure, delivery, return) over consecutive days.
  4. Multi-Perspective Data Collection:
     • People: Intercepting and interviewing individuals encountered along the robot's path (office workers, residents, sanitation workers, etc.).
     • Places: Observing the physical environment from the robot's eye level (approx. 29 inches) to identify obstacles like tactile paving, drainage grates, and low-hanging banners.
     • Time: Analyzing how interactions change based on temporal rhythms (e.g., rush hour vs. lunch breaks).

3. Key Contributions

• Methodological Innovation: WawR provides a structured framework for conducting ethnography with autonomous robots in "in-the-wild" settings, moving beyond ad-hoc observation.
• Shift in Agency: It repositions the robot from a passive object of study to an active "guide" that determines the research field, boundaries, and trajectory.
• Bridging Temporal Disconnects: By walking with the robot, researchers can seamlessly transition from observation to interviewing, solving the common ethnographic problem of the time gap between observing an event and interviewing a participant.
• Holistic Scope: The method expands the scope of analysis to include non-human actors (animals, infrastructure, weather) and the "mundane" aspects of robot existence that are often overlooked.

4. Key Results & Findings

The methodology was applied in two field sites in Seoul, South Korea (a high-density commercial mall area and a mixed residential/office district). Key findings include:

• Asymmetrical Interaction Norms: Humans consistently yield to robots. Robots rarely initiate avoidance maneuvers; instead, they use auditory cues (horn or voice: "Hello, I'm delivering. Let me pass by") to command space, and humans accommodate them.
• Absence of Personal Space with Robots: Unlike interactions with humans, pedestrians frequently touch robots, fiddle with antennas, or even open food compartments without permission, indicating a lack of established social norms for human-robot physical boundaries.
• Robot-Centric Environmental Insights:
  • Robots often take routes not visible on standard GPS maps (e.g., specific shortcuts).
  • Obstacles at "robot eye level" (e.g., uneven pavement, low banners) significantly impact mobility but are often invisible to human researchers.
  • Robots do not honk for non-human obstacles (e.g., dogs, boxes) or other robots, only for humans.
• Temporal Dynamics:
  • Perception Shifts: Office workers viewed robots as nuisances during lunch rushes but found them "endearing" or amusing during quieter times.
  • Routine Integration: Over time, robots became "invisible" to regular workers and residents, blending into the daily rhythm of the neighborhood, whereas researchers initially expected high novelty effects.
• Social Networks: The method revealed hidden relationships, such as a café owner who had pre-existing ties with the robot company, influencing the robot's route and frequency of visits.

5. Significance

• For HRI Research: WawR offers a robust protocol for studying commercial robots that cannot be controlled by researchers. It moves the field from "lab-based" or "short-term pilot" studies to long-term, naturalistic ethnography.
• For Urban Planning & Design: By highlighting how robots interact with infrastructure (e.g., drainage grates, curb heights) and social norms, the findings provide critical data for designing better urban environments for AMRs.
• Theoretical Impact: The paper validates the "more-than-human" perspective in robotics, demonstrating that understanding the robot's experience (its "wheels") is essential to understanding the social dynamics of the city.
• Practical Application: The paper provides a step-by-step guide and practical tips (e.g., using smartphones for note-taking to avoid suspicion, working in teams) for other researchers attempting similar field studies.

In conclusion, the WawR methodology transforms the robot from a subject of observation into a co-researcher that guides the investigation, offering a deeper, more nuanced understanding of how autonomous systems integrate into the complex fabric of urban life.