A humanoid can look convincing in a prepared demonstration. A pilot has a higher bar: it evaluates a task in realistic conditions and provides evidence for another investment, a change in scope, or stopping the project.
1. Define the task and the baseline
- Describe the task steps, start and end conditions, required inputs, and expected outcome.
- Describe the environment, space, lighting, surfaces, people, and nearby movement.
- Record how the task is done now, including time, interruptions, exceptions, and required skills.
- Define permitted use, explicit restrictions, and situations that stop a trial.
2. Choose measurable acceptance questions
Agree in advance how many attempts will be run and which environmental variations will be included. Set the sample and conditions for the project and its risks; a universal minimum without knowing the task would be misleading. Record every attempt consistently and distinguish task completion from manual help, resets, or technician intervention.
For each attempt, record:
- Did the robot complete the task under the agreed definition of success?
- How long did the full cycle take, including setup, waiting, reset, and human intervention?
- When and why did a person need to pause the task or take over?
- How sensitive was the result to changes in the part, position, surroundings, or input data?
- What connections, tools, software, or workplace changes did the solution require?
3. Agree supervision, safety, and responsibilities
Name the customer and manufacturer contacts, the person authorized to stop operation, and who can approve a restart after a fault or update. Define how errors are reported, access to the equipment and data, responsibility for integration, and the technical support scope. Risk assessment must match the specific place, task, and system configuration; this checklist does not replace it.
4. Evaluate the whole solution and decide
Include integration, tools, space changes, training, human supervision, maintenance, downtime, and support alongside the platform price. Before starting, define which outcomes mean continue, change the task, or stop the pilot. Base that decision on recorded evidence and agreed scope, not an impressive video.
The final record
- Objective, location, duration, platform configuration, and any changes made.
- Success definition, attempt plan, outcomes, and failed cases.
- Required supervision, interventions, downtime, and safety findings.
- Total costs and unresolved technical or operating questions.
- The customer decision and conditions for any next phase.
A well-designed pilot can conclude that a humanoid is not yet suitable for the task. That result is useful when it is supported by evidence, described openly, and helps choose a better next step.
Sources and further reading
- Humanoid robots in the MALONI portfolio ↗MALONI Robotics
- Regulation (EU) 2023/1230 on machinery ↗EUR-Lex
- Collaborating robots: risk assessment and workplace design ↗European Agency for Safety and Health at Work
Manufacturer specifications are attributed in the article. Safety references link to the relevant European sources. This article is practical orientation, not a substitute for an application-specific engineering or legal assessment.