This is the direction where being wrong hurts someone, so it is written more plainly than the rest.
The most expensive misunderstanding in the field
The phrase "collaborative robot" is routinely read as "robot that is safe to work next to". That reading has caused real injuries.
What a collaborative-rated robot provides is a set of capabilities: force and power limiting, sensitive collision detection, and the ability to operate in modes that support sharing space with people.
What it does not provide is a safe application.
The robot is assessed by its manufacturer. The application is assessed by whoever installs it, and that is where the safety actually lives.
Three ways a collaborative robot becomes dangerous.
The end effector. Force limiting in the arm says nothing about a sharp tool, a hot component, a welding torch or a gripper that can crush a finger. The hazard moved to the thing on the end.
The workpiece. A robot carrying something sharp, heavy, hot or chemically hazardous presents that hazard regardless of its own force limits.
Trapping and geometry. A limited-force robot can still pin a person against a fixture, a wall or another machine. Force limits assume free space, and a trapped body part does not have any.
So the correct statement is that collaborative capability enables applications that may be made safe after assessment. It never certifies one.
What the direction covers
The scope: how people and robots share a workspace, interaction, trust and working safely side by side.
Four areas.
Risk assessment. Identifying hazards, estimating risk, selecting protective measures and verifying they work. The core activity and a structured discipline with defined method.
Protective measures. Guarding, interlocks, light curtains, scanners, safe speed and separation monitoring, force and power limiting, and emergency stop systems.
Functional safety. Where a safety function is implemented in control systems, the reliability of that function has to be designed, quantified and verified against a required level.
Human factors. Predictability, signalling intent, trust calibration and foreseeable misuse, which is the part that determines whether protective measures survive contact with real work.
The four modes of collaborative operation
Worth knowing, because procurement conversations go better when both sides use the same vocabulary.
Safety-rated monitored stop. The robot stops when a person enters the shared space and resumes when they leave. Simple and widely used.
Hand guiding. An operator moves the robot directly using a control device. Useful for teaching and for assisted lifting.
Speed and separation monitoring. Sensors track the person's position and the robot slows or stops as separation decreases. Powerful and dependent on sensing reliability and on stopping distance calculations that must account for the robot's actual dynamics with its actual payload.
Power and force limiting. The robot is designed so that contact does not exceed defined limits. This is the mode most people mean by "collaborative", and it is the one where end effector and workpiece hazards most often invalidate the assumption.
These modes can be combined, and the appropriate combination comes out of the risk assessment rather than out of a product brochure.
Where this sits in the domain
Human-robot interaction and robot safety is the ninth and final direction in Astra Trainer's robotics and autonomous systems domain, and it applies to all eight before it: industrial cells, service robots in public spaces, autonomous vehicles and drones each carry their own version of the same question.
Partners with robots on a floor typically scope it with industrial robotics and with the advanced manufacturing domain, where quality engineering and reliability, and maintenance and asset management cover the procedures the safety case depends on. Lessons are five minutes, which matters here because the people who most need this content are maintenance and production staff who cannot leave the floor. You can see the nine directions here.
Where people actually get hurt
The pattern in robot-related incidents is consistent and it is not what people expect.
Serious incidents predominantly occur during maintenance, teaching, programming, setup and fault clearance, rather than during normal automatic operation.
The reason is structural. During normal operation the protective measures are active and working. During those other activities, someone is inside the space with the guarding bypassed or the robot in a mode where normal protection is reduced.
Four specific patterns.
Clearing a jam without isolating. A stopped robot is not a de-energised robot. It may resume, or a stored energy source may release.
Unexpected restart. Someone else resets, or a signal arrives, or a program resumes from a different point than expected.
Unexpected motion path. The robot moves somewhere the person did not anticipate, which is more likely during teaching and after a program change.
Reduced-speed complacency. Working in a teach mode at reduced speed, treating it as safe. Reduced speed reduces severity and does not eliminate it, and a robot at reduced speed can still trap and crush.
The workforce conclusion is direct: the people who most need robot safety competence are maintenance technicians, programmers and production staff, not just the engineer who designed the cell.
Why this capability is scarce
Three reasons, all structural.
It spans disciplines. Machine safety requires mechanical understanding, controls and functional safety, plus regulatory knowledge, plus the judgement to assess a real application. That is the same cross-domain problem as mechatronics.
It is only visible when it fails. A well-assessed installation produces no events, so the capability shows up as an absence.
It is frequently treated as documentation. Risk assessment gets produced to satisfy an auditor rather than to design the system, which is how you end up with a compliant file and an unsafe cell.
Against that, the route is unusually well defined. Machine safety has recognised competence frameworks and formal qualification schemes in several jurisdictions, which makes it one of the clearer career progressions available to a technician.
The roles, named
Machinery safety engineers and consultants.
Functional safety engineers. Designing and verifying safety-related control functions to a required performance level.
Risk assessment specialists for automation projects.
Safety validation engineers. Verifying that protective measures perform as designed, including stopping distance measurement.
Human factors specialists for shared workspaces.
Compliance and CE or UKCA marking specialists for machinery and assemblies.
Health and safety professionals with machinery competence, which is a narrower group than general health and safety.
Maintenance supervisors responsible for safe systems of work around robots.
Who can be trained into it
Maintenance technicians. The most important group. They are the ones inside the cell, and they carry the risk personally. Formal machine safety competence is also a genuine progression route, with recognised qualifications attached.
Controls engineers. Into functional safety, which is the natural technical extension of what they already do.
Health and safety professionals. Understand the regulatory framework and assessment method, and frequently lack the machinery and controls depth to assess an automation application properly.
Automation and project engineers. Need safety competence to design rather than retrofit it, which is always cheaper.
Production supervisors. Into safe systems of work, since they control how tasks are actually performed.
Quality engineers. Already work with structured assessment and verification, which transfers.
This article is not a risk assessment and cannot be used as one. Robot installations are subject to machinery safety law, and the obligation to carry out a risk assessment and apply adequate protective measures sits with the party placing the system into service. Requirements, standards and their status differ by jurisdiction and change over time. Safe systems of work, energy isolation procedures and authorisation to enter a robot cell are site-specific and must be established locally by competent people. Structured training builds understanding and supports recognised competence routes. It does not confer machine safety competence certification, functional safety certification, or authority to assess, approve or sign off any installation.
What to take from this
Safety is a property of the application. A collaborative-rated arm carrying a sharp tool, a hot part or a heavy load is not a safe system, and trapping defeats force limits.
Four collaborative modes exist and the right combination comes from the risk assessment, not from the product description.
People get hurt during maintenance, teaching and fault clearance, when protection is reduced, rather than during normal operation.
Reduced speed lowers severity and does not remove the hazard, and treating it as safe is the most common complacency in the field.
And the people who most need this competence are the maintenance technicians who go inside the cell, for whom it is also one of the clearest qualification routes available.
Are collaborative robots safe to work alongside?
Not automatically. Collaborative rating describes the robot's capabilities. Safety is a property of the whole application, and end effectors, workpieces and trapping geometry can all make a force-limited robot dangerous.
When do robot injuries happen?
Predominantly during maintenance, teaching, programming and fault clearance, when someone is inside the space with normal protective measures reduced or bypassed, rather than during automatic operation.
Is reduced speed enough protection?
No. It reduces severity and does not eliminate the hazard. A robot at reduced speed can still trap and crush, and treating teach mode as inherently safe is a common and dangerous assumption.
Who needs robot safety training most?
Maintenance technicians, programmers and production staff, because they are the people who enter the cell. Not only the engineer who designed it.
Is there a career route in machine safety?
Yes, and an unusually clear one. Machine safety has recognised competence frameworks and formal qualification schemes, which makes it a genuine progression for maintenance and controls staff.