Collaborative Robots
Cobot vs Industrial Robot: What Actually Differs
A cobot is an industrial robot. The difference sits in the risk assessment, not in the arm, and since February 2025 both are covered by the same standard.
A cobot is an industrial robot. Both fall under the same safety standard, ISO 10218, and the only thing that separates them is whether the risk assessment permits a human to share the working space while the arm moves. That is a property of the application, not of the machine, and it is why a cobot behind a fence is legal, while a cobot without a fence may still be illegal.
One standard, two marketing words
The vocabulary is the source of most confusion. There is no legal category called "collaborative robot". ISO 10218-1 covers the robot itself, ISO 10218-2 covers the integrated application, and both were revised in February 2025, the first full rewrite since 2011. The revision did something that settles the argument: it pulled the content of ISO/TS 15066:2016, the technical specification everybody quotes when they say "cobot standard", directly into ISO 10218-2:2025. The technical specification no longer exists on its own.
What the standard actually regulates is collaborative operation, and it names four types. A machine sold as a cobot may run in any of them, in more than one during a single cycle, or in none at all.
| Type | What the robot does | Human may enter | Typical use |
|---|---|---|---|
| Safety-rated monitored stop | Holds position with power on, motion stops before entry | Yes, while stopped | Machine loading, inspection |
| Hand guiding | Moves only through a hand-operated enabling device | Yes, in contact | Teaching, lifting assistance |
| Speed and separation monitoring | Slows or stops as the measured gap closes | Yes, at distance | Shared cells, part handover |
| Power and force limiting | Contact is allowed and capped by design | Yes, contact permitted | Assembly, screwdriving, dispensing |
Only the fourth type actually permits the robot to touch a person while moving. The first three are separation strategies dressed in different clothes. A large six-axis robot with a payload of 500 kg can legally run in safety-rated monitored stop, and it is then operating collaboratively even though nobody would ever call it a cobot.
What genuinely differs in the hardware
Machines marketed as cobots do share a recognisable engineering profile, and it is worth naming precisely, because these traits are what make power and force limiting feasible rather than merely aspirational.
- Torque sensing in the joints. Contact is detected in single-digit milliseconds by the joint itself rather than inferred from motor current, which is what allows a reaction before the force limit is exceeded.
- Rounded geometry and low surface pressure. Pressure limits in the standard run from 110 N/cm² at the face to 300 N/cm² at the fingertip. A sharp edge fails the pressure test long before it fails the force test.
- Low moving mass. Effective mass drives the permissible speed. The standard's own worked example shows a robot with 1 kg effective mass allowed 2,400 mm/s against a hand, while 20 kg drops that to 1,900 mm/s, and against a chest the same 20 kg allows only 400 mm/s.
- Modest payload and reach. Commercial cobot arms cluster between 3 kg and 30 kg payload with reaches of roughly 500 mm to 1,750 mm, whereas industrial arms run from 3 kg to over 2,300 kg.
What does not differ: repeatability. Cobot datasheets typically quote ±0.02 mm to ±0.1 mm, which sits in the same band as comparable industrial arms. Anyone who assumes a cobot is imprecise is reading the wrong specification.
The trade that decides the project
Speed is the honest difference. Power and force limiting caps the tool centre point velocity at whatever the contact model permits, and that number is often four to eight times below the arm's mechanical maximum. An industrial arm behind a fence routinely runs at 2,000 mm/s to 3,000 mm/s and beyond; the same task in contact-permitted operation may be capped near 400 mm/s if the exposed body region is the chest.
When the collaborative route wins anyway
Collaborative operation pays where floor space, changeover frequency or part variety dominates the economics rather than raw cycle time.
| Situation | Better choice | Reason |
|---|---|---|
| Single product, 3-shift operation, cycle-bound | Fenced industrial arm | Full speed, guarding cost amortises in weeks |
| Batch sizes under 500, weekly changeover | Collaborative operation | Redeployment without re-guarding |
| Floor space below 4 m² per station | Collaborative operation | No fence footprint, no door swing |
| Payload above 35 kg | Fenced industrial arm | Contact limits become impractical |
| Sharp tools, welding, laser | Fenced industrial arm | The tool is the hazard, not the arm |
| Machine tending with operator present | Safety-rated monitored stop | Full speed while alone, stop on entry |
Note the last row. It is the most under-used configuration in the whole catalogue: a full-speed industrial robot that simply stops in a safety-rated way when a person steps in. It gives away nothing in cycle time and still removes the fence door and its interlock queue.
The paperwork does not care what you bought
Whichever arm is on the bench, the integrator carries the same obligations. A risk assessment is required, the safety functions need a rated performance level, and the completed cell needs a declaration of conformity before it is switched on for production. Buying a machine that the vendor calls collaborative removes none of that. In the European Union the rules tighten on 20 January 2027, when Regulation (EU) 2023/1230 replaces Directive 2006/42/EC for machinery placed on the market, with no overlap period between the two.
The practical consequence: budget for validation measurement. Force and pressure at every reachable contact point have to be measured with a calibrated instrument against the limits in the standard, not assumed from the vendor's brochure. A cell with 12 reachable contact scenarios is a day of measurement, and it is the step most often discovered late.
Frequently asked questions
Is a cobot legally a different machine from an industrial robot?
No. Both are industrial robots under ISO 10218. The standard regulates collaborative operation, which is a property of the application and its risk assessment, not a class of machine. ISO 10218-1:2025 does introduce two robot classes, but they are defined by hazard potential, not by the word cobot.
Does ISO/TS 15066 still apply?
Its content does, its status does not. The February 2025 revision folded the power and force limiting requirements and the biomechanical limits of ISO/TS 15066:2016 into ISO 10218-2:2025, and the technical specification no longer stands alone. The Annex A limit values themselves were not weakened.
Can a cobot run without a safety fence?
Only if the risk assessment for that specific application says so. The permission comes from the assessment and the validation measurements, not from the purchase order. A cobot carrying a sharp tool, a hot part or a heavy workpiece frequently ends up behind guarding anyway, because the hazard belongs to the end effector.
How much throughput does collaborative operation cost?
Expect a factor of two to eight on the moving segments of the cycle, depending on which body region is exposed and how much effective mass the arm carries. The standard's worked example allows 2,400 mm/s against a hand at 1 kg effective mass but only 400 mm/s against a chest at 20 kg.
Are cobots less accurate than industrial robots?
No. Typical cobot repeatability is quoted between ±0.02 mm and ±0.1 mm, the same band as comparable industrial arms. Accuracy, which is a different specification, is usually worse than repeatability on both types unless the robot has been calibrated.
What payload is realistic for collaborative operation?
Commercial collaborative arms run roughly from 3 kg to 30 kg. Above about 35 kg the permissible contact speeds fall so low that the cycle stops being economic, and separation-based operation or fixed guarding is the practical answer.
Sources
- ISO 10218-1:2025, Robotics, Safety requirements, Part 1: Industrial robotsInternational Organization for Standardization, published February 2025
- Updated ISO 10218: answers to frequently asked questionsAssociation for Advancing Automation (A3), on the incorporation of ISO/TS 15066 and the new robot classes
- ISO/TS 15066:2016, Annex A, biomechanical limits and transient contact speed tablesForce, pressure and speed limit values quoted in this article
- Regulation (EU) 2023/1230 on machineryEU-OSHA legislation record, application date 20 January 2027