The HRI Field
Human-Robot Interaction (HRI) is the study of how humans and robots communicate, collaborate, and affect each other. HRI spans a spectrum from fully teleoperated systems (human in control at all times) to fully autonomous robots (human sets goals, robot executes independently), with shared autonomy and collaborative robotics (cobotics) occupying the middle ground.
HRI research draws on robotics engineering, cognitive science, psychology, and human factors. Key questions include: How should a robot communicate its intentions to nearby humans? How do humans build (or lose) trust in robot systems? What interaction modalities — voice, gesture, gaze, force — work best in which contexts?
Collaborative Robots (Cobots)
Collaborative robots are industrial manipulators specifically designed to work safely alongside human operators without fixed safety fencing. They contrast with traditional industrial robots that must be caged because their speed and force would cause injury on human contact.
ISO/TS 15066 defines four collaborative operation modes:
- Safety-rated monitored stop: robot stops whenever a human enters the collaborative workspace
- Hand guiding: operator physically guides the robot by grasping it; robot follows with zero-force control
- Speed and separation monitoring: robot slows or stops as human approaches, resuming when distance is safe
- Power and force limiting (PFL): robot limits contact forces so that accidental collision causes no injury — the defining feature of cobots
Leading cobot manufacturers: Universal Robots (UR3e–UR20), FANUC CR series, ABB YuMi / GoFa, KUKA LBR iisy, Techman TM, Doosan.
Key cobot features: rounded covers, no pinch points, built-in force/torque sensing, power limiting firmware, easy hand-teaching (move-and-teach), and app-based deployment (UR+ ecosystem).
Cobot Platform Comparison
| Platform | Payload | Reach | TCP Force Limit | Standout Feature |
|---|---|---|---|---|
| Universal Robots UR10e | 12.5 kg | 1300 mm | Configurable (ISO 10218) | Largest UR, embedded F/T sensor |
| ABB YuMi (IRB 14000) | 0.5 kg / arm | 559 mm | PFL certified | Dual-arm, padded arms, no guarding needed |
| FANUC CR-35iA | 35 kg | 1813 mm | Soft-stop on contact | Heaviest payload certified collaborative robot |
| KUKA LBR iisy 11 | 11 kg | 1300 mm | Joint torque sensors | Torque-sensitive joints, 7-DoF |
| Techman TM12 | 12 kg | 1300 mm | PFL certified | Integrated eye-in-hand vision system |
Social Robots and Non-Industrial HRI
Beyond the factory floor, robots increasingly interact with people in social contexts: hospitals, schools, retail, eldercare, and homes. Social robots must navigate conversational norms, express and interpret non-verbal cues (gaze direction, body posture, facial expression), and maintain appropriate personal space (proxemics).
Pepper (SoftBank Robotics): humanoid torso on a wheeled base with tablet screen, cameras, and microphones — deployed as a receptionist and customer service agent worldwide.
NAO: small humanoid used extensively in education and autism therapy research — its predictable, non-threatening form facilitates interaction with children.
PARO: therapeutic robotic seal for dementia patients — designed purely for emotional interaction, shown in clinical trials to reduce anxiety and medication use.
Key HRI design principles:
- Legibility: robot motions should make the robot's intentions obvious to nearby humans
- Predictability: consistent behaviour builds trust; surprising movements undermine it
- Appropriate anthropomorphism: enough human-like cues to feel natural; avoiding the Uncanny Valley — the discomfort caused by robots that are almost-but-not-quite human
- Natural language interaction: LLM-powered dialogue (GPT-4, Claude) dramatically lowers the barrier to robot interaction for non-technical users
Trust and Acceptance in HRI
- 01
Trust is built through consistent, predictable, and transparent robot behaviour over repeated interactions
- 02
Over-trust (automation bias) is as dangerous as under-trust — humans may not monitor a robot closely enough
- 03
The Technology Acceptance Model (TAM) predicts adoption based on perceived usefulness and perceived ease of use
- 04
Explainability matters: robots that can articulate why they made a decision are trusted more than black-box systems
- 05
Failure mode design: robots that fail gracefully (safe stops, clear error messages) preserve trust; unexpected failures destroy it
- 06
Cultural differences significantly affect preferred robot appearance, interaction style, and acceptable personal space