A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Collaborative robots (cobots) work alongside humans in shared workspaces, requiring fundamentally different safety, control, and interface design than traditional industrial robots that are isolated behind fences. Safety in HRI is ensured through force/torque limits (robots stop if forces exceed thresholds to prevent injury), compliant motion control (forces are bounded), quick emergency stop systems, and speed/power limits. Beyond safety, effective HRI requires intuitive interfaces (demonstrative learning, gesture recognition, natural language commands), understanding human intent (learning from human corrections), and predicting human actions to avoid collisions. HRI is an interdisciplinary field spanning robotics, ergonomics, cognitive science, and HCI. Applications include manufacturing (assembly, inspection), healthcare (surgery assistance, rehabilitation), and domestic service (care robots, household assistants). The challenge is building robots that are safe, predictable, and easy for non-experts to operate and supervise.
For decades, industrial robots operated in cages, separated from human workers by fences and interlocks. Robots were fast, powerful, and dangerous — a collision could break bones or cause crush injuries. Humans entered only when the robot was safely parked and locked. But this model doesn't scale to factories with space constraints, flexible manufacturing, and tasks requiring human judgment and dexterity alongside robotic precision. Collaborative robots (cobots) promise to change this: robots working alongside humans, sharing workspaces, and collaborating on tasks.
Safety as a Core Design Principle: Collaborative robots cannot achieve safety through physical separation — the whole point is shared workspaces. Safety must be built into the robot's interaction with humans. ISO/TS 15066 and related standards specify maximum force and power limits for different types of contact. For a collision with the robot's arm: max 220 N. For sustained contact with the hand: max 140 W power. These limits are determined from biomechanics studies of human injury thresholds. Exceeding them can cause bruising, fractures, or severe trauma. Cobots are designed mechanically and controlled to respect these limits.
Mechanical Compliance: Cobots use lighter materials, smaller actuators, and flexible joints compared to traditional industrial robots. Some designs include series elastic actuators (SEA): a spring between the motor and the end-effector allows force measurement and compliance. Others use purely software compliance via impedance control. The goal is that a cobot "feels soft" when touched — it yields, not rigidly resists.
Control Strategies for Safety: Force-limiting control is central. The robot monitors forces and torques at the end-effector and joints. If forces exceed safe limits, power is cut or reduced immediately. The reaction is fast (millisecond-scale) but not instantaneous — momentum carries the robot forward briefly. To prevent injury, safe cobots also limit speed in collaborative zones. A robot moving at 0.2 m/s has much less momentum than one moving at 2 m/s; limiting speed reduces injury risk. Modern cobots often operate in collaborative zones at 0.25-0.5 m/s, much slower than traditional robots (1-2 m/s) but sufficient for many assembly tasks.
Impedance Control for Collaboration: Cobots often use impedance control: the robot behaves like a virtual spring with controlled stiffness. The robot has a desired trajectory (from a learned task or programmed path), but if forces develop (e.g., a human pushes the robot's arm), the robot yields according to compliance laws. This makes the cobot "feel" responsive to human interaction — if a human guides the cobot's arm, it follows naturally. If a part resists insertion, the robot's force feedback causes it to adjust position or reduce pushing force. Impedance control is why cobots are sometimes called compliant — they don't rigidly push through resistance; they adapt.
Intent Prediction and Situation Awareness: Cobots can improve safety and efficiency by predicting human intent. If the robot knows the human will reach toward location A, it can move out of the way or prepare to assist. Intent prediction uses human gaze direction, hand position, activity recognition (what task is being performed), and learned patterns of human behavior. However, predictions are imperfect. Conservative safety design responds to prediction uncertainty by assuming the worst: if the robot doesn't confidently predict the human's action, it assumes the human might reach toward the robot and moves away. This causes over-reactivity but prevents collisions.
Design Principles for HRI Safety:
1. Accessibility of emergency stops: Red buttons, clearly visible, reachable from any position
2. Operational modes: Clear distinction between teach/demo mode (robot moves slowly) and run mode (faster but still limited), with mode indication
3. Predictability: Robots move in anticipated directions, accelerate/decelerate smoothly (no jerky motion)
4. Communication: LED indicators show robot status; sounds signal robot motion
5. Speed limits: Velocity capped in collaborative zones; faster speeds only in human-free areas
6. Force limits: Active monitoring and control ensure interaction forces stay safe
Applications and Limitations: Cobots excel at assembly, inspection, loading/unloading, and collaborative tasks. They're increasingly used in healthcare (surgery assistance where the cobot constrains motion to safe bounds) and manufacturing. But they're slower than traditional robots and less powerful, limiting applications requiring high speed or force. A car assembly line with multiple fast-moving tasks may be impractical for cobots. A small electronics assembly line with flexible routing and quality inspection is ideal.
The Future: As cobots become more dexterous and intelligent (better sensing, learning), collaborative applications will expand. Autonomous cobots that understand context (which part goes where, what the human needs next) and communicate (via gestures, AR, natural language) will blur the line between tool and teammate. But the safety-first design philosophy — force limits, compliance, predictability — will remain central.