A cobot that stops when someone walks up to it is not automatically safe. In an industrial setting, collaborative robot safety requirements are determined by the application, the tooling, the workpiece, the surrounding machine design and the people interacting with it. That is why two cells using the same robot arm can have very different safety measures.
For engineers, OEMs and plant teams, the key issue is straightforward. A collaborative robot is not a shortcut around machine safety obligations. It is a different design approach that can reduce guarding in some tasks, but only when the risks have been properly assessed and controlled. If the end effector is sharp, the payload is unstable, or the task involves trapping points, full collaboration may not be suitable at all.
What collaborative robot safety requirements actually cover
Collaborative robot safety requirements sit across several layers. There is the robot’s built-in functional safety capability, the cell-level safety design, the application-specific risk assessment and the operating procedures around use, maintenance and changeover. Treating the robot’s data sheet as the whole answer is where many projects come unstuck.
In practice, the safety case must consider how the robot moves, how people approach it, what happens when contact occurs and whether other hazards remain present even at reduced speed. A cobot may be designed for human interaction, but that does not remove hazards created by grippers, fixtures, product edges, conveyors or adjacent equipment.
The standards framework matters here. Most project teams will be looking at ISO 10218 for industrial robot safety, ISO/TS 15066 for collaborative applications, and the broader machinery safety expectations that apply to the equipment as a whole. In Australia, those requirements need to be interpreted in the context of local WHS duties and the specific site environment. Compliance is not only about selecting certified hardware. It is about demonstrating that the complete machine or cell is safe for its intended use.
The four collaborative modes and why they matter
A lot of confusion comes from using the word collaborative too broadly. Not every cobot application involves direct physical interaction. The recognised collaborative approaches generally fall into four modes, and each has different safety implications.
Safety-rated monitored stop is the simplest. The robot stops when a person enters the collaborative space, and motion only resumes when conditions are safe. This can suit load and unload tasks where people need periodic access, but it is not the same as working side by side while the robot is moving.
Hand guiding allows an operator to physically guide the robot through motion, usually for teaching or assisted handling. The safety focus here is on enabling devices, controlled motion and preventing unexpected behaviour.
Speed and separation monitoring is more dynamic. The robot continues operating while maintaining a safe distance from a person, typically using safety sensors or vision systems to monitor approach. This can support better productivity, but the validation is more demanding because detection performance, stopping time and approach speed all affect the protective distance.
Power and force limiting is what most people picture when they think of a cobot. Here, the robot is designed so that contact forces remain within acceptable limits for the body region exposed. Even then, the application still needs assessment. A low-force robot fitted with a rigid sharp tool or handling a hard-edged component can still create unacceptable risk.
Risk assessment is the real starting point
Before selecting barriers, scanners or a robot model, the job is to complete a proper risk assessment. That means identifying every reasonably foreseeable hazard during normal production, setup, cleaning, maintenance, fault recovery and restart.
For collaborative applications, the assessment should look closely at crushing, trapping, impact, shearing and puncture hazards. It also needs to consider how operators actually behave. If staff will naturally reach into the process to clear misfeeds or realign product, that access pattern must be part of the design. A cell that is safe only when everyone follows a perfect sequence on paper is usually not safe enough in service.
The end effector deserves special attention. In many real installations, the robot arm is not the primary risk. The hazard comes from a vacuum tool dropping product, a mechanical gripper pinching fingers, or a spindle, blade or hot component mounted to the flange. Payload shape and stiffness also matter. A lightweight robot handling a large awkward item can present more risk than the arm itself suggests.
This is also where integrators need to decide whether collaboration is genuinely justified. Sometimes a compact guarded cell with well-planned access is the better engineering outcome. It may be easier to validate, easier to maintain and more productive over the life of the asset.
Key design controls for collaborative robot safety requirements
Once the risk assessment is done, the required controls become clearer. These often include safety-rated monitored functions inside the robot controller, external safety devices and mechanical design measures.
Safe limited speed, safe stop, safe position and safe monitored motion are common functional safety tools, but they must be configured to suit the application. Stopping performance must be measured, not assumed. If a scanner is used to trigger reduced speed or stop zones, the detection fields need to reflect real approach paths and not just the neat geometry shown in the layout drawing.
Mechanical layout still does a lot of safety work. Rounded fixtures, controlled pinch clearances, limited stroke, workpiece presentation and sensible operator standing positions can reduce risk before software logic is even considered. That usually produces a more reliable result than trying to solve every hazard through sensing alone.
Where collaboration is only required for part of the process, hybrid designs often make sense. A cell may use guarding for high-risk steps and allow monitored operator access in lower-risk phases. This is a practical way to balance throughput and safety without forcing the whole machine into a single operating concept.
Validation, documentation and change control
A collaborative cell is not finished when it powers up and runs parts. Safety validation is essential. This includes checking safety functions, proving stopping times, verifying sensor coverage, confirming mode transitions and testing foreseeable fault conditions.
Documentation should be clear enough for maintenance teams and future project engineers to understand how the safety concept works. That includes limits, assumptions and any restrictions on tooling, payload or speed. If a line team later swaps the gripper, increases product mass or changes the fixture height, the original safety assessment may no longer hold.
That point is often underestimated in brownfield sites. Collaborative applications are sensitive to change because the acceptable force, speed and separation parameters are linked to the exact task. Good change control is therefore part of meeting collaborative robot safety requirements, not an administrative extra.
Common mistakes in collaborative applications
The most common mistake is assuming a collaborative robot removes the need for guarding. In reality, many cobot cells still need fixed guards, interlocked access, area scanners or light curtains. Collaboration can reduce guarding in the right task, but it rarely removes protective measures entirely.
Another issue is focusing only on normal operation. Fault clearing, manual jog, tool change and cleaning can introduce higher risk than automatic mode. If those states are not designed properly, the system may be compliant on paper but awkward and unsafe in day-to-day use.
There is also a tendency to overestimate productivity gains. Collaborative operation can improve flexibility and reduce footprint, but speed limits may reduce cycle time performance. For some applications, especially in packaging, materials handling or machine tending, the trade-off is worthwhile. In higher-throughput or higher-energy processes, a traditional guarded industrial robot may still be the better fit.
A practical path for project teams
For OEMs, integrators and end users, the best approach is to define the task first and then decide whether collaboration is appropriate. Start with the workpiece, the required motion, the operator interaction and the production target. Then complete the risk assessment and identify which collaborative mode, if any, is suitable.
From there, select safety-capable hardware that supports the required functions and validate the whole cell as a machine, not just the robot. Local technical support helps here because many problems are application problems rather than catalogue problems. Suppliers that understand motion, sensing, safety and integration can help avoid overdesign in one area and under-protection in another.
For businesses specifying or upgrading robotic systems, this is where practical engineering input matters. Tech Source works with industrial customers that need more than part numbers, particularly where robotics, safety devices and control architecture need to work together in a compliant and maintainable way.
Collaborative robotics can deliver real value, but only when the safety concept is matched to the job rather than the marketing. The right question is not whether a robot is collaborative. It is whether the complete application can be made safe, practical and productive for the people who will use it every day.