Proxemics

This zone is reserved for whispering, touching, and close physical contact. For robots, intrusion is typically highly inappropriate and perceived as a severe threat or intimacy violation.

• HRI Application: Strictly avoided in most social contexts. Permissible only for specific physical tasks (e.g., a healthcare robot administering a sensor) with explicit user consent and clear signaling.
• Safety Imperative: Requires Power and Force Limiting (PFL) safety modes if contact is possible.
• Example: A robot reaching across a user's body to grab a tool without warning would violate this zone, causing alarm.

The bubble maintained for conversations with friends and family. This is the primary zone for collaborative work and fluid HRI.

• HRI Application: The optimal range for a collaborative robot (cobot) sharing a workbench. The robot can hand over tools, receive parts, and observe human actions.
• Interaction Design: Robots should approach this zone deliberately, often using gaze estimation or verbal cues to seek implicit permission before entering.
• Example: A fetch robot delivering a package stops just within this zone, allowing the human to comfortably reach out and take it.

The distance for impersonal business and social gatherings. This zone is for observation and non-intrusive presence.

• HRI Application: Ideal for social navigation by mobile robots in hallways or open offices. Robots can observe human activity for intent recognition or action anticipation from a comfortable distance.
• Monitoring & Approach: Robots in this zone are perceived as available but not imposing. They may use this zone to stage before moving into the personal zone for interaction.
• Example: A guide robot waiting near a museum exhibit entrance, allowing visitors to approach it first.

The distance for public speaking or performances. At this range, fine-grained interaction is not expected.

• HRI Application: Used for long-range human motion forecasting and trajectory planning for autonomous vehicles or robots in large, dynamic spaces like warehouses or airports.
• Function: The robot is part of the environment backdrop. Its primary goal is to be predictable and avoid inadvertently entering closer zones without cause.
• Example: An autonomous mobile robot (AMR) plotting a path across a factory floor, using this zone to plan around groups of workers from a distance.

Proxemics is not just about distance; the relative orientation and approach vector are equally critical for signaling intent.

• Frontal Approach: Direct and engaging, but can be confrontational if too rapid. Used for initiating deliberate interaction.
• Oblique/Side Approach: Perceived as less threatening and more natural for joining an ongoing activity or moving alongside a human.
• Facing Direction: Robots should generally orient their "front" towards the human during interaction to signal attention, mimicking human facing formation in conversation.

Proxemic norms are not universal; they vary significantly by culture, individual preference, and context.

• Cultural Differences: Acceptable conversation distances are smaller in Latin American and Middle Eastern cultures than in Northern European or East Asian cultures. A robot must adapt its social navigation parameters accordingly.
• Contextual Factors: The appropriate zone shrinks in crowded environments (elevators, public transport) and expands in formal or dangerous settings.
• HRI Design Implication: Requires adaptive systems that can perceive context and, where possible, learn individual user preferences over time to avoid causing proxemic anxiety.

Robots use proxemic zones to plan paths that are socially compliant. This involves:

• Maintaining a respectful distance when passing or following a person.
• Adhering to cultural norms like passing on a specific side.
• Adjusting speed and trajectory to avoid proxemic violations that cause discomfort.

Example: A delivery robot in a hospital corridor slows and yields space when approaching a group of people, rather than cutting through them.

For collaborative robots (cobots), proxemics informs safe spatial configurations. It dictates:

• Optimal interaction distances for joint tasks (e.g., within the social zone for hand-over tasks).
• Safe approach vectors and orientations that are non-threatening.
• Integration with safety standards like ISO/TS 15066, where proxemic zones can trigger Speed and Separation Monitoring (SSM).

This ensures fluid human-robot teaming without cages.

Robots in public spaces (hotels, airports, retail) use proxemics for approach behavior and service initiation.

• A robot determines if a person is in a public (can approach) or intimate (do not approach) zone.
• It uses gaze estimation and body orientation to assess engagement before initiating interaction.
• It manages turn-taking in conversations by modulating distance and posture.

This prevents the robot from being perceived as intrusive or annoying.

In therapy, education, and elder care, Socially Assistive Robots (SAR) leverage proxemics to build rapport and provide effective support.

• A robot therapist may start at a social distance and slowly decrease to a personal distance as trust is built.
• It mirrors a user's proxemic behavior to establish affective connection.
• It respects heightened intimate zones for users with sensory sensitivities.

Proper spatial management is critical for trust calibration and therapeutic outcomes.

Proxemics guides where and how a robot should position itself relative to a human for effective communication. Key considerations include:

• F-Formation theory: Positioning within a person's o-space (conversational space) at an appropriate angle for interaction.
• Avoiding positions directly behind (threatening) or too far in front (distant).
• Aligning the robot's "face" or frontal sensors with the human's orientation to signal attention.

This is fundamental for natural language interaction and gesture recognition.

In scenarios with multiple humans, robots use proxemics to understand group dynamics and intervene appropriately.

• Detecting proxemic clusters to identify conversational groups.
• Inferring social relationships based on interpersonal distances (e.g., intimate vs. social).
• Planning approaches that respect group o-spaces and do not disrupt the interaction.

This requires advanced scene understanding and integrates with Theory of Mind (ToM) concepts to reason about social attention.

Social Navigation is the algorithmic planning and execution of robot paths that adhere to implicit human social norms within shared environments. Unlike basic obstacle avoidance, it incorporates proxemic principles to ensure a robot's movement is predictable, courteous, and non-disruptive.

Key considerations include:

• Passing conventions (e.g., keeping to the right in certain cultures)
• Overtaking protocols and signaling intent
• Dynamic personal space bubbles that adjust for context (e.g., a hallway vs. an open plaza)
• Gaze and trajectory prediction to avoid cutting across a human's intended path

Intent Recognition is the computational process by which a robot infers a human's immediate goals or planned actions from multimodal observations. It is a critical precursor to appropriate proxemic behavior, as spatial norms change based on what a human is trying to do.

A robot might infer intent from:

• Trajectory and gait analysis (e.g., walking toward a door vs. milling about)
• Object affordances (reaching for a tool suggests a task)
• Contextual cues (standing at a counter implies a service interaction)
• Prior interaction history

Accurate intent recognition allows a robot to anticipate whether to approach, maintain distance, or yield space.

Human Motion Forecasting is the task of predicting the future trajectory or body pose sequence of a person based on their observed past motion. This is a foundational capability for proactive proxemics, enabling a robot to adjust its own position before a human enters its intimate space.

Modern approaches use:

• Temporal convolutional networks or recurrent neural networks to model motion history
• Social pooling layers to account for interactions between multiple people
• Graph neural networks to model the kinematic structure of the human body

Predictions are typically short-term (1-3 seconds) but are essential for fluent, collision-free cohabitation of dynamic spaces.

Power and Force Limiting (PFL) is a collaborative robot safety mode defined in ISO/TS 15066 where the robot's inherent design or control system limits the power and force of its movements. PFL is the enabling safety standard for close-proxemics interaction, as it ensures any incidental contact will not cause injury.

Implementation involves:

• Force/torque sensors in joints to monitor exerted energy
• Compliant actuators or control strategies (e.g., impedance control)
• Biomechanical thresholds for pain and injury (e.g., for specific body regions)
• Rounded, padded physical design to minimize pressure

PFL allows robots to operate in the personal (45-120 cm) and occasionally intimate (0-45 cm) zones defined by proxemics, which would be prohibited with traditional industrial robots.

Affective Computing is the interdisciplinary study and development of systems that can recognize, interpret, and simulate human emotions. In proxemics, a person's emotional state directly modulates their preferred interpersonal distance (e.g., agitated individuals may require more space).

A robot with affective computing capabilities could:

• Use facial expression analysis or vocal tone analysis to infer emotional valence
• Adjust its approach distance and orientation based on perceived stress or openness
• Modify its own communicative behavior (e.g., speech rate, lighting) to de-escalate
• This creates a feedback loop where proxemic behavior is dynamically tuned to the human's affective state, promoting comfort and trust.

Theory of Mind (ToM) in AI refers to an agent's capacity to attribute mental states—such as beliefs, intents, desires, and knowledge—to others. For proxemics, this means a robot must not only observe where a human is, but also model what the human knows and intends about the shared space.

A robot with a functional ToM might reason:

• "This human has not seen me, so they do not know I am moving behind them." → Robot should announce presence.
• "This human believes this chair is occupied." → Robot should not path through that perceived space.
• "This human desires privacy in this conversation." → Robot should increase its approach distance.

This higher-order reasoning moves proxemics from simple distance-keeping to truly socially intelligent spatial behavior.