ISO/TS 15066 and collaborative robot operation

What the document is

ISO/TS 15066 is a technical specification, not a standard. The distinction is procedural and it matters: a technical specification is published when a committee has useful guidance but has not reached the consensus or maturity required for a full International Standard. It was developed to supplement the ISO 10218 pair, which covers industrial robots and robot systems, by giving concrete guidance on one topic those documents treated only in outline — operation in which a person and a robot share a workspace.

It is a guidance document for people performing a risk assessment. It is not a certificate, it is not a conformity mark, and no robot is approved by it.

There is no such thing as a collaborative robot

This is the single most useful thing to understand about the subject, and the phrase "collaborative robot" works against it.

Safety is a property of an application, not of a machine. The same arm, with the same firmware, is safe in one cell and unsafe in another, because what makes it safe or unsafe is the task, the tooling, the workpiece, the speed, the layout and who is standing where. A robot built with features that support collaborative operation — torque sensing in each joint, rounded surfaces, limited mass — still becomes dangerous the moment you fit a blade to it, or a hot part, or run it fast enough past a pinch point.

So a vendor can honestly say a robot is designed for collaborative applications. Nobody can honestly say an application is safe without a risk assessment of that application. The guide on risk assessment covers how that process actually runs.

The four modes of collaborative operation

The specification describes four ways a person and a robot can share space. They are not a hierarchy and they are frequently combined within one cell, with different modes active in different zones or at different points in a cycle.

• Safety-rated monitored stop. The robot holds position with drives energised while a person is in the shared space, and resumes when the space is clear. The person and the moving robot are never in the space at the same time. This is the simplest mode to justify and the most widely deployed.
• Hand guiding. An operator moves the robot directly using a device designed for the purpose, with an enabling control. The robot moves only while it is being guided. Common in teaching and in assisted lifting.
• Speed and separation monitoring. The robot and the person may both move in the shared space, with a sensing system maintaining a protective separation distance that varies with the robot's speed and the person's approach. Reduce the distance and the robot slows; close further and it stops. This is the mode that enables genuine concurrent work, and it is the most demanding to engineer because the sensing has to be as trustworthy as the stop it triggers.
• Power and force limiting. Contact between the robot and a person is permitted, with the energy of that contact constrained by design and by control so that it stays below limits established for the body region that could be contacted. This is the mode people usually mean when they say collaborative, and it is the most frequently misunderstood.

Power and force limiting, without the numbers

The last mode is the one that needs care, because it is the one where the document contains specific biomechanical values and where summarising those values secondhand would be irresponsible.

The principle: research on pain onset across different parts of the human body was used to derive limits on the force and pressure that a contact event may produce, differentiated by body region, and distinguishing a transient contact from a sustained or clamping one. An application designed around this mode has to show, for each contact that is reasonably foreseeable, that the energy delivered stays within the applicable limit.

What that requires in practice: identifying which body regions can be contacted, in what manner, at what point in the cycle; establishing the robot's speed, mass and the geometry of whatever it is carrying; and then measuring, with instrumentation designed for the purpose, rather than calculating from a datasheet. The measured values are compared against the limits in the document itself.

This guide does not reproduce those limits, and you should be wary of any source that does. They are differentiated by body region in ways that a summary destroys, they changed between document versions, and an application designed against a number remembered from a blog post is an application designed against nothing. If you are doing this work, the document is the source, and a competent safety engineer reads it with you.

Where the specification now sits

Two things are true at the same time here, and collapsing them into one is the commonest error in secondary writing on this subject.

The subject matter moved. ISO 10218-1:2025 cancels and replaces the 2011 second edition, and it incorporates safety requirements for industrial robots used in collaborative applications — material that previously lived in this technical specification.

The specification itself has not been withdrawn. ISO’s catalogue record still lists ISO/TS 15066:2016 as published. Its status is given as an International Standard to be revised, with that revision opened in June 2025, and a successor numbered ISO/AWI 15066-1 is listed as under development.

So do not write, and do not accept, that TS 15066 has been replaced or superseded. Its topics have been incorporated into a newer standard while the document remains published and under revision. Which document governs a particular application, in a particular jurisdiction, at a particular edition, is a question for a qualified safety assessor working from the documents themselves rather than from this page.

Checked 14 November 2026 against the ISO catalogue record for ISO/TS 15066:2016. Check it again before relying on it — a status of "to be revised" is precisely the kind of fact that changes.

The guide on robot safety standards sets out how the documents in this area relate to one another.

What this means if you are buying

A few questions separate a serious conversation from a marketing one.

• Which mode or modes does the proposed application use, and in which zones?
• Who is performing the risk assessment, and is that person independent of the party selling the cell?
• If power and force limiting is proposed, who is measuring the contact events, with what equipment, and will you see the results?
• What happens to the assessment when the tooling, the workpiece or the cycle time changes — because one of those will change?
• Does the end effector or the workpiece introduce a hazard the robot itself does not have? Sharp, hot, heavy or energised changes the answer entirely.

Not a compliance document

This page is an explanation written for people trying to understand the landscape. It is not legal or safety advice, it is not a compliance document, and it does not establish that any system meets any requirement. Conformity is determined by a competent assessor against the applicable documents in your jurisdiction, working from the documents themselves.

Four misreadings that cause real problems

"It is a collaborative robot, so we do not need guarding." The assessment decides what guarding is needed, not the product category. Plenty of applications built on arms designed for collaborative operation end up fenced, because the task made them dangerous.

"Power and force limiting means it cannot hurt anyone." It means contact energy is constrained below limits derived from pain onset, for identified body regions, in identified contact scenarios. It says nothing about a sharp tool, a hot part, a pinch point against a fixed surface, or a region nobody assessed. Crushing against a hard surface is a different scenario from free contact, and it is treated differently.

"The robot is certified." Components carry conformity. Applications are assessed. A certificate in a vendor's brochure describes the machine they sold you, not the cell you built with it.

"We did the assessment at commissioning." The assessment is tied to the application as it existed that day. Change the end effector, the workpiece, the speed or the layout and the basis of the assessment has changed with it. Teams that treat the assessment as a one-time gate rather than a living document are the ones who get surprised.

Mobile changes the question

Everything above assumes a robot that stays where it is bolted. A mobile robot, or an arm mounted on a mobile base, is a different problem: the workspace is not fixed, the people in it were not briefed, and the relevant documents are not the same ones. Industrial trucks and automated guided vehicles have their own standards, and a mobile manipulator sits awkwardly between families that were written separately.

If your application is mobile, the first question to your safety engineer is which documents apply at all, before any discussion of modes. That is a genuinely unsettled area and it is not one to work out from a summary page.