Collaborative vs. Non-Collaborative Robots: Safety Rules and Risks Compared (OSHA)

In OSHA's Technical Manual, a collaborative industrial robot application is one designed for direct interaction with workers; a non-collaborative application is designed without that need and is typically separated from workers by traditional machine safeguarding. The distinction is about the application, not the robot's brand label, and it changes how hazards are controlled. Either way the risk is real: NIOSH identified 41 robot-related fatalities in the U.S. between 1992 and 2017.

The definitions OSHA uses

The OSHA Technical Manual, Section IV, Chapter 4, was written to help compliance officers and others who inspect facilities with robot systems. It divides industrial robot applications ‘based on the degree of interfacing capability with workers’.

• Collaborative: designed for direct interaction with workers. OSHA notes that other definitions treat robot-to-robot interaction as collaborative, but the Manual requires direct interaction with workers; robot-to-robot applications are classed as non-collaborative.
• Non-collaborative: all other types and structures of industrial robots, designed without the need for direct interaction with workers and typically separated from them by traditional machine safeguarding.
• Industrial mobile robots (IMRs): can be either. They navigate autonomously to automate transport tasks, use obstacle or collision avoidance, are not intended to carry a driver, and can carry a manipulator to move between applications.

NIOSH's overview frames robot types slightly differently: traditional industrial robots work in cells and cages away from workers; professional service robots assist workers; and collaborative robots directly interact with people, a group in which NIOSH includes wearable robotics or powered exoskeletons and co-existing or mobile robots.

Side-by-side comparison

How non-collaborative cells are safeguarded

OSHA's approach is separation during automatic operation. In most circumstances the application should automatically reach a safe state when a worker enters the safeguarded space. When someone must work with the robot still active, typically for programming or teaching, safety rests on three things: manual mode, an enabling device (usually a 3-position device held in the center-on position, often built into the teach pendant), and reduced speed.

Administrative controls complete the picture: written entry and exit procedures and training, lockout/tagout SOPs and training under 29 CFR 1910.147, and PPE under Subpart I. OSHA stresses that these come after attempts to design out the hazard and to safeguard have been exhausted.

How collaborative applications are made safe

A collaborative application uses one or more of these technologies in automatic mode, as described by OSHA:

• Speed and separation monitoring (SSM): presence-sensing detects intrusion; at minimum the robot stops, and some integrations slow down first and stop before contact. Where speed is used for safety, a safety function should monitor that it is not exceeded.
• Hand-guided controls (HGC): the robot moves under the worker's direct control, for example lifting a heavy box while the worker holds a hold-to-run control.
• Power and force limiting (PFL): contact is expected and permitted only when forces and pressures are limited so there is no injury, through inherently safe design or control means such as joint torque sensors.
• Safety-rated monitored stop (SMS): power to actuators is retained while the robot is held at standstill; OSHA notes it is not used alone but with SSM, HGC or PFL.

OSHA also explains that SSM is commonly combined with PFL so the robot can run fast when nobody is nearby, then slow to speeds at which contact would be permissible. For PFL, the risk assessment must identify the expected contact areas on the body and whether contact is transient (the body part can move away) or quasi-static (trapped against a fixture), then compare measured pressures and forces with permissible biomechanical limits. Those limits are set out in RIA TR R15.606-2016, the U.S. adoption of ISO/TS 15066:2016, a paid document whose values are not reproduced here. OSHA adds that contact with sensitive body regions such as the face, temples and throat is to be prevented.

Do you really need a collaborative application?

OSHA suggests answering these questions before choosing collaborative operation:

1. Is the presence of a person integral to the application?
2. Do the robot and person have to share a workstation?
3. Do they have to work on the same workpiece simultaneously?
4. Have task locations been identified and made known, with safe access?
5. Does the person need physical contact with the robot, end-effector or workpiece while it moves?
6. Is the robot and end-effector designed for collaborative use, and does the application have the needed safety functions?

A light robot is not automatically harmless. The Technical Manual warns that even low-energy robots with payloads as low as 6-1/2 pounds (3 kilograms) can be used in very dangerous applications, which is why the application, including the end-effector and workpiece, is what gets classified and assessed.

Whichever category you choose, OSHA expects the safeguards to be verified, not trusted. Its Technical Manual says site acceptance testing should be performed by the integrator and verified by the user before initial start-up, and that employers should then periodically check stopping performance, safety distances and safety-function settings; checking the checksums of safety parameters is described as a quick way to see whether settings have changed since the last inspection.

What the injury data says

NIOSH lists struck-by or caught-between hazards, crushing and trapping, slips, trips and falls, and electrical hazards among robot risks, and flags emerging risks for robots working close to people, including unexpected contact, distraction and mental stress. Exposure is growing: per the NIOSH overview, the number of industrial robots in U.S. factories grew by 10% in 2022, and service robots sold for professional use in the U.S. rose to 158,000 units in 2022, a 48% increase.

• OSHA's Technical Manual cites studies from Sweden and Japan indicating many robot accidents happen not in normal operation but during assembly, installation, testing, programming and maintenance.
• NIOSH FACE reports referenced by OSHA include a 29-year-old mold setter killed when struck on the head by a cycling single-side gantry robot, and a 23-year-old carousel operator killed when his foot tripped a light sensor and a computer-controlled robotic platform came down on him.
• A typical documented failure in non-collaborative cells: perimeter guarding that let a worker enter the safeguarded space without causing a protective stop.

To check a definition or figure against the source, ask the robot and cobot safety base, for example ‘What makes a robot application collaborative versus non-collaborative in OSHA's framework?’