Human-robot collaboration is a mode of operation in which people and robots share a workspace and, often, a task, with safety ensured by safety-rated robot functions and a documented risk assessment rather than by full physical separation. In robot safety, it is the discipline of deciding how close a person may come, how fast the robot may move and how hard it may touch.
The term describes an application, not a machine. The same robot can run fenced in one cell and collaboratively in another, and a cell is collaborative only when its design, safeguards and validation allow intended interaction between a person and moving robot parts.
Levels of interaction between people and robots
Human-robot collaboration covers a spectrum of interaction depth, from people merely passing near a robot to joint handling of the same part. Each level raises the demands on safety functions and validation.
| Level | Shared workspace | Shared time | Example |
|---|---|---|---|
| Coexistence | No, separate zones | Yes | Robot and worker in adjacent unfenced zones |
| Sequential cooperation | Yes | No, they take turns | Worker loads a fixture, robot then welds |
| Simultaneous cooperation | Yes | Yes, on different steps | Robot screws while worker inserts cables |
| Responsive collaboration | Yes | Yes, on the same part | Robot holds a panel while worker fastens it |
Most industrial installations operate at sequential or simultaneous cooperation. Responsive collaboration, in which the robot adapts to human motion in real time, is technically demanding and remains less common in production.
Safety functions that make collaboration possible
Collaborative applications rely on safety-rated functions executed by a safety controller, with integrity typically specified as performance level d with category 3 architecture under ISO 13849-1, or an equivalent safety integrity level. These functions differ from ordinary control features because a single fault must not cause their loss.
- Safety-rated monitored stop: the robot holds still, with motion monitored, while a person is inside the zone.
- Speed and separation monitoring: sensors measure distance and the robot slows or stops before contact becomes possible.
- Power and force limiting: contact is allowed but forces and pressures stay below body-region limits.
- Hand guiding: the operator moves the robot directly, using an enabling device.
- Safe zones and tool limits: configurable boundaries keep the tool and robot away from defined areas.
Speed and separation monitoring is often designed with the protective separation distance concept from ISO 13855 and ISO/TS 15066. A simplified form is S = K × T + C, where K is the approach speed of a person, commonly taken as 1,600 mm/s for walking, T is the total stopping time of detection plus robot braking, and C is an allowance for intrusion and measurement uncertainty.
Risk assessment for collaborative applications
Every collaborative application begins with a risk assessment following the method of ISO 12100: identify hazards, estimate and evaluate risk, then reduce it in a fixed order of inherently safe design, safeguarding and information for use. ISO 10218-2 adds requirements specific to robot cells.
Collaborative cells introduce hazards that guarded cells avoid. The table below lists typical ones together with common risk reduction measures.
| Hazard | Where it arises | Typical measure |
|---|---|---|
| Transient contact | Robot moving freely strikes an arm or torso | Power and force limiting, rounded tooling |
| Quasi-static clamping | Body part trapped between robot and fixed object | Clearance gaps, lower force limits, reversing |
| Sharp or hot tooling | Gripper edges, knives, welding tips | Covers, stop before tool reaches the person |
| Dropped parts | Grip lost during motion | Self-locking grippers, limited lift heights |
| Ergonomic strain | Awkward postures at the shared station | Height adjustment, task redistribution |
Clamping is usually the decisive hazard, because the body cannot recoil. Limits for quasi-static contact are considerably stricter than for transient contact, which is why layouts avoid placing hands or heads between the robot and rigid surfaces.
Human-robot collaboration compared with other safety concepts
Human-robot collaboration sits alongside other robot safety concepts, and confusing them leads to unsafe or overly cautious designs. The comparison below shows their distinct roles.
| Concept | Role | Relation to collaboration |
|---|---|---|
| Human-robot collaboration | Permits intended interaction | The overall operating mode |
| Safeguarding by separation | Keeps people out with guards | Alternative to collaboration |
| Emergency stop | Complementary protective measure | Required, but not a collaboration method |
| Safety controller | Executes safety-rated functions | Hardware that enables collaboration |
| Risk assessment | Determines required measures | Mandatory basis for every collaborative cell |
An emergency stop is not a collaborative safeguard. It depends on a person noticing danger and reacting, whereas collaborative functions must act automatically before harm occurs.
Validation and common mistakes
A collaborative application is validated by measurement, not by datasheet claims. Integrators measure contact forces and pressures with calibrated instruments at the relevant body regions, check stopping times and distances, and verify that every safety function behaves as specified after configuration.
- Assuming a cobot makes the cell safe, regardless of tooling and task.
- Validating once and never again, even after programs, speeds or grippers change.
- Using default safety parameters instead of values derived from the risk assessment.
- Ignoring foreseeable misuse, such as reaching into the cell from an unexpected side.
- Underestimating stopping distance at higher speeds or with heavier payloads.
Safety parameters should be protected by password and checksum in the safety controller, and any change should trigger a documented revalidation.
Stripped to the essentials
Human-robot collaboration means people and robots working side by side. No fence keeps them apart. The robot does some of the work and the person does the rest. Sometimes they even handle the same object.
To make this safe, engineers first study every way a person could get hurt. Then they set rules for the robot. It must slow down when someone comes near. It must never push harder than a safe limit. It must stop at once when something unexpected happens. Finally, the team measures and tests these rules before anyone works in the cell.
Where it is applied
A manufacturer of electrical cabinets assembles heavy door panels. Previously two workers lifted each panel while a third fastened the hinges. The company introduces a collaborative application in which a robot lifts and holds the panel in position while a single worker aligns it and drives the screws.
The risk assessment identifies clamping between the panel and the cabinet frame as the main hazard. The integrator limits robot speed to a low value near the frame, adds a laser scanner for speed and separation monitoring on the approach side and measures forces at the hand and forearm. The result is one fewer person per station, less lifting strain and a documented validation that is repeated whenever the program changes.