Working with Cobots: Collaborative Robot Risks

Why does cobot safety not mean "no fence equals safe"?

A conventional industrial robot is separated from employees by a fence and an interlocked door; when someone enters the cell, the robot stops. A collaborative robot, commonly called a cobot, is designed to remove this separation: while an employee places a part, the robot beside them can tighten screws, pack or feed a machine. In automotive supplier and metalworking workshops in Bursa and neighbouring provinces, cobots are most often found in machine feeding, screwdriving, adhesive application, palletising and quality control stations.

The most common misconception about cobot safety is the idea that "we bought a cobot, so no fence is needed". The robot manufacturer provides the arm's force and speed limiting functions; but it does not know what is fitted to the end of the arm, at what speed it is run, or which part of the employee's body could come into contact under which conditions. This is why international standards treat collaboration as a feature of the application, not the robot. In the 2025 editions of ISO 10218-1 and ISO 10218-2, most of the content on collaborative applications from ISO/TS 15066, previously a separate technical specification, was moved into ISO 10218-2.

For the rules in fenced robot cells, see our article on industrial robot cell safety; this article focuses on the specific risks of unfenced, shared work.

Four modes of collaborative robot operation

In practice, collaborative work is set up in four modes. More than one can be used together at the same station.

Mode of operation	How does it work?	Typical use	Point to watch
Safety-rated monitored stop	When the employee enters the shared space the robot stops and the stop is monitored; when the employee leaves it resumes where it left off	Part loading and unloading	If the stop is not monitored, the robot may move unexpectedly
Hand guiding	The employee guides the arm by hand using a control handle	Positioning heavy parts, teaching	Enabling switch and emergency stop on the handle
Speed and separation monitoring	Sensors monitor the employee's distance; the robot slows as they approach and stops at a set distance	Large shared areas, palletising	Sensor blind spots, wrong placement
Power and force limiting	The robot may make contact, but contact force and pressure are limited	Assembly, screwdriving, side-by-side work	Trapping, sharp and hot tools, high speed

Which mode is chosen depends on whether contact must be prevented or limited. In any situation where contact is unacceptable, for example where a cutting tool is used, force limiting is not relied upon; contact is prevented by speed and separation monitoring or by physical separation.

Power and force limiting: what contact is acceptable?

In power and force limiting applications, two types of contact are distinguished:

Transient contact: The robot strikes the employee, but the employee's body can recoil. For example, a moving arm touching the shoulder.
Quasi-static (clamping) contact: A body part is caught between the robot and a fixed object, such as a table, fixture, wall or machine body. The pressure continues until the arm stops.

Clamping contact is far more dangerous than transient contact, because the force does not dissipate to a place where the body can escape. Most incidents in which cobot collision risk turns into serious injury are of this type. The first task in station design is therefore not to leave trapping points in the arm's path: fixtures are positioned with a gap between the arm and the table that will not trap a hand, and trajectories where the arm approaches a wall or column are changed.

Other points to watch in power and force limiting:

The body region matters. Standards set different limits for regions such as the hand and arm than for the head and neck. In practice, contact with the head and face is treated as something to be prevented by design; the robot's trajectory is kept away from the employee's head height.
Speed determines force. The same arm produces much greater impact energy at high speed. Increasing speed "to shorten cycle time" can invalidate the application's safety assessment.
No decision without measurement. Contact force and pressure are measured at the contact points with a suitable measuring device after installation is complete, and recorded. The force limit entered in the software may not match the actual measured value.
The payload is taken into account. The mass and edges of the part carried by the arm change the contact force and the risk of cuts.

Speed and separation monitoring: is the area scanner in the right place?

In speed and separation monitoring, the aim is for the robot to stop before the employee reaches it. The protective separation distance required for this is determined by calculating together the robot's stopping time and distance, the employee's approach speed, the sensor's response time and measurement uncertainty. Common mistakes on site are:

Placing the area scanner in a spot that "looks right" without measuring the robot's stopping distance.
The scanner leaving a blind spot behind a pallet, crate or column.
The stopping distance increasing when the robot starts working with a heavier tool or part, but the protective field not being updated.
Scanner fields being reduced or bypassed so that production is not disrupted.

The protective field must be revalidated after every change and must only be changeable by an authorised person. Whether the safety functions achieve the required reliability, that is, the performance level, must be shown in the integrator's technical file.

Cobot collision risk: the tool and part more than the robot

The collaborative robot arm itself is designed with rounded contours and soft surfaces. Most of the risk comes from the end effector fitted to the end of the arm:

Grippers and vacuum heads: Can trap fingers; may release and drop the part when energy is cut.
Cutting and drilling tools: A blade, drill bit or deburring tool causes injuries that force limiting cannot protect against.
Hot tools: Adhesive guns, soldering irons or welding torches bring a burn risk.
Screwdriving units: The rotating bit and torque can catch a glove or finger.
The part carried: Sharp-edged sheet metal, glass or heavy castings increase the severity of contact.

The end effector is therefore treated as a separate line in the station's risk assessment. For principles relating to conventional machine guards, see our article on machine guards.

Cobot risk assessment step by step

A cobot risk assessment starts before the robot is purchased and does not end at commissioning. The recommended steps are:

Define the application. The part, tool, cycle, the employee's task and the limits of the shared space are written down.
List the contact scenarios. During which movement, which part of the body, against which object could contact occur? Transient and clamping contacts are distinguished.
Choose the mode of operation. If contact is unacceptable, speed and separation monitoring or physical separation; if acceptable, power and force limiting.
Eliminate trapping points by design. Fixture layout, trajectory and speed are adjusted accordingly.
Measure and validate. Contact forces are measured, scanner fields and stopping times are tested, and the emergency stop is tried.
Document and train. The results are entered in the risk assessment; operators and maintenance staff are trained in the application-specific rules.
Manage change. A new product, new tool, speed increase or station relocation reopens the assessment. For this, see our article on management of change.

On the legislative side, the Machinery Safety Regulation (2023/1230/EU), which applies from 20 January 2027, also covers partly completed machinery. A robot arm sold on its own usually falls into this group and is supplied by its manufacturer with a declaration of incorporation and assembly instructions. When the arm is combined with the end effector and peripheral equipment into a working application, who will ensure the conformity of the whole must be stated clearly with the integrator in the purchase contract. For details, see our article on the Machinery Safety Regulation and CE. For the employer, the responsibility to select equipment suitable for the work and use it safely continues under the Regulation on Health and Safety Requirements for the Use of Work Equipment.

The employee's role in daily use

Keeping a cobot safe depends on small habits on site:

At the start of each shift, the emergency stop button and the area scanner are checked to be working.
Program, speed and force settings are changed only by an authorised person and with a record.
Every unexpected contact, even without injury, is reported as a near miss; this report triggers a review of the force setting or trajectory.
Before a new table, crate or rack is added to the station, check whether it creates a trapping point.

We carry out the risk assessment of stations with cobots and the analysis of contact scenarios as part of our risk assessment service. For other risks in the automotive supply industry, see our automotive supply industry page.

Frequently Asked Questions

Is a fence needed when using a cobot?

It depends on the application. In an application with validated power and force limiting or speed and separation monitoring, a fence may not be needed. However, in applications involving cutting, hot or heavy parts, physical separation is often still required.

Does power and force limiting protect against every contact?

No. It is effective in transient impacts; it is not sufficient on its own for clamping contact, the head and face region, or cutting or hot tools. These situations are resolved by design or by preventing contact.

Does increasing a cobot's speed affect the risk assessment?

Yes. As speed increases, contact force and stopping distance change. After a change in speed, tool or trajectory, force measurements and protective fields must be revalidated.

Who carries out the cobot risk assessment?

The integrator usually carries out the design and conformity work for the application. The risk assessment covering the conditions of use in the workplace is carried out by the employer's risk assessment team and combined with the integrator's documents.