Ergonomics and humanoid robotics workshop for factory teams: RULA/REBA/NIOSH and prioritizing high-risk tasks

How the brief came about

The aim of this workshop was to choose which risky jobs to fix first using figures, not feelings.

MSDs (Musculoskeletal Disorders, meaning work-related injuries to muscles and bones) do not happen in a single day like accidents; they build up over years before symptoms appear. The costs are therefore invisible at first: sick leave, lower productivity, loss of skilled workers, and the expense of having to rework production lines later.

This course also answers the question of which jobs Humanoid robots (robots shaped like people) and Cobots (robot arms that can work alongside people) should be used for. The course's answer starts with data on the job, and only then chooses the solution.

Requirements and constraints

We designed this course to finish in 1 day and to let participants start measuring jobs without needing a budget yet. The requirements and constraints used in designing the course were as follows.

• Time: 1 day, divided into 2 parts. The first part covers the problem and measurement; the second covers handing work over to robots.
• Participants came from different companies and were divided into 5 groups, so activities had to use jobs that each person knew from their own workplace.
• The final activity's worksheet allowed participants to estimate frequency themselves if they did not yet have real figures, and to give a rough risk score of 1–10. The camera data was a case study in the slides, not data that participants used in the activities.
• In 2026, Humanoid robots are still almost entirely pilot projects, so the course also offers options that do not require buying a robot.
• When people and robots work in the same area, the safety standards ISO 10218 and ISO/TS 15066 must be referenced.

Why this workshop has you measure jobs before considering robots

Because choosing which job to fix requires 2 figures: frequency and risk. Without these two figures, choosing a robot or equipment is guesswork. The course therefore sets out 3 steps: measure so the problem is visible, fix it through engineering, then consider robots only for jobs that remain risky and frequent enough.

Tools such as RULA, REBA and NIOSH are already reliable. The problem lies in the data fed into them. If the frequency figure comes from an estimate, the assessment result is an estimate too. With figures from actual counting, the engineering team can show management clearly which job to invest in fixing first.

1. 1

   Identify jobs where employees complain of aches and pains

   In the first activity, each person writes down jobs or workstations from their own workplace where employees often complain of aches and pains, where it hurts, and how many times per hour the job is done. This list is used again in the final activity.

2. 2

   Try assessing with the naked eye first

   Each group looks at pictures of two kinds of job, a bus driver and a street sweeper, then scores the risk by eye alone. This activity shows that judging by feel gives inconsistent answers.

3. 3

   Learn the standard tools

   RULA scores upper-body posture. REBA covers the whole body, including the legs and balance. The NIOSH Lifting Equation calculates the recommended lifting weight from reach distance, height, lifting distance, twisting, frequency and type of grip, then compares it with the actual weight as the Lifting Index (LI).

4. 4

   Review a case study of AI cameras collecting the figures

   In the case study presented, AI Vision cameras (cameras that analyse images with AI) time each step and count actual work cycles throughout the shift without disturbing operators. The counted frequency values are used in the NIOSH equation and to support RULA scoring, instead of estimates from supervisors.

5. 5

   Place jobs on the matrix and choose which to fix first

   In the final activity, each group chooses 2–3 jobs from the list on the board, estimates their frequency, gives a rough risk score of 1–10, then places them on the matrix in the table below.

Deliverables

We delivered a 1-day course designed entirely in-house and served as trainer throughout the day. The deliverables were as follows.

• The "Humanoid and Ergonomics for Modern Industry" course and slides, in 2 parts
• 3 sets of activity worksheets: identifying jobs employees complain about, scoring risky postures with the naked eye, and placing jobs on the matrix
• A case study of using AI Vision cameras to count actual work cycles throughout the shift
• Safety precautions when people and robots work together, in line with ISO 10218 and ISO/TS 15066
• Steps for starting to measure at real workstations, which can be done without a budget yet, as the closing content of the workshop

Confirmed results from participants' worksheets

The confirmed results of this case are the worksheets submitted by participants. All 5 participant groups submitted worksheets for Activity 2 and the final activity on the day of the training. In Activity 2, some groups saw street sweeping as riskier, while others saw bus driving as riskier.

In the final activity, each group placed jobs on the frequency and risk matrix to choose which to fix first. The figures in the worksheets were values estimated by participants themselves, not values measured by cameras. We have no information on whether each team took the steps to measure at real workstations after the training, so we do not report post-training results.

What we learned

This work offers 3 principles for choosing which risky jobs to fix first, and for running training on this subject.

• Prioritising jobs requires a single set of criteria. In Activity 2, the groups looked at the same set of job pictures yet chose different jobs as riskier. Scoring by eye therefore cannot be used to compare jobs; tools such as RULA, REBA or NIOSH are needed.
• An assessment result is only as reliable as the data fed into it. The figures in the worksheets this time were values estimated by participants themselves. Estimates suit practising the way of thinking in a training room, but actual work cycles should be counted before making an investment decision.
• Activity exercises should come from a list of real jobs from one factory. Participants this time came from different companies, and some groups placed bus driving and office work on the matrix. When running training inside a factory, prepare the factory's job list in advance, so that the matrix results can be put to further use.

How this approach can apply to your factory

You can start the same approach without a budget yet. Begin with a workstation where people often complain of aches and pains, then measure the job before choosing a solution. The table below separates the reusable parts from those that need adapting to the nature of the job and what your factory already has.

• Suited to factories with lifting, material handling or repetitive same-posture work throughout the shift, such as packing departments, warehouses or assembly lines
• Participants should come from engineering or job design, not just HR, because most fixes lie in the workstation, equipment, frequency and weight
• Prepare in advance a list of jobs or workstations where employees often complain of aches and pains, to use as activity exercises
• If cameras are used to observe working posture, use the data to adjust workstations, not to assess individual employees, and inform employees before installation