Start with a clearly defined pilot task and the intended workspace: a convincing demonstration does not prove that a robot can perform the task in your workplace. Buying may make sense if your team wants to develop in-house over several projects and can verify access to the interfaces it needs. For a limited trial, a supported rental may be an option, while simulation can help you explore workflows in advance but cannot replace testing real hardware.
If employees may be in the robot’s range of movement, teams need to assess the specific workspace, potential collisions and appropriate protective measures. The DGUV Institute for Occupational Safety and Health and the Federal Institute for Occupational Safety and Health explain the requirements for workplace risk assessments. However, general guidance on cobots cannot replace an assessment of the specific humanoid robot and its intended location.
These teams want to explore a specific industrial task first and, if appropriate, test it using their own hardware. Kempten University of Applied Sciences describes research projects using demonstrators for industry partners; Unitree Robotics, Booster Robotics and Fourier document development tools for their respective platforms.
If you are planning a narrowly defined trial, you can compare buying and providing ongoing in-house support with a supported rental solution. X-Tech Robotics advertises rental, commissioning and support; the services included and who is responsible for what must be agreed contractually for the specific trial.
A good fit if: your team wants to develop control, perception or manipulation in-house over several trial cycles, and the necessary interfaces are available for the specific model variant.
Consider: In-house access to the hardware brings integration, maintenance and support responsibilities. An SDK proves neither that the robot is suitable for the task nor that it is safe for the workplace.
A good fit if: you want to trial a clearly defined task using real hardware for a limited period, and can agree support arrangements and responsibilities.
Consider: Development access, costs, insurance and responsibility for protective measures are not guaranteed simply because rental and support are offered.
A good fit if: movement sequences or software integration are still being worked out and you want to explore them first without employees working alongside moving hardware.
Consider: Simulation can help you explore workflows, but it cannot prove that gripping or protective measures will work in the conditions of the real workplace.
For industrial humanoid research and pilot projects in Germany, the key questions when choosing depend on the task, development access, the effort involved and the intended workspace. The figures below group consumers by their main priority.
| Priority | What to check before buying | Who it matters most to |
|---|---|---|
| Safety in the workspace | Clarify the robot’s range of movement and operating modes, as well as the planned workflow’s stopping and protective functions. | If employees might come close to moving hardware, I want to be able to assess risks in the specific workplace. Around 29% of consumers put safety first. |
| Feasibility of the pilot task | Define the task, objects, routes and measurable success criteria; consider what the humanoid form is intended to achieve compared with a specialised solution. | Anyone wanting to test a clearly defined industrial task needs a credible feasibility trial, not just a demonstration. For around 29% of consumers, task feasibility is the main priority. |
| Development access | Confirm which interfaces, control permissions and software integrations are available for the exact model variant. | If my team wants to develop its own control systems or movement sequences, access must match the platform supplied. Around 24% of consumers prioritise development access. |
| Effort and responsibilities | Set out services and responsibilities for integration, support, maintenance, spare parts and insurance in writing. | Anyone planning a limited trial will want to understand ongoing tasks and responsibilities early on. Around 18% of consumers put predictable overall costs first. |
A research demonstrator does not automatically prove that the workflow will work with different objects, routes or operating conditions in your workplace.
Define the task, test objects and measurable success criteria before choosing a platform or trial approach.
Different objects and workflows may require additional development work and further trials.
Manufacturers document development tools, but the interfaces and permissions available may depend on the specific model variant.
Ask for written confirmation of the model variant, interfaces and usage rights for your planned trial.
An unsuitable configuration may require additional integration work.
Neither a rental agreement nor a digital demonstrator automatically assesses the hazards involved in using the actual hardware.
For a real-world trial, plan for the workplace risk assessment and checks on the agreed protective measures. Under § 3 BetrSichV, employers must assess risks before work equipment is used.
Workplace design, testing and technical support should be included as separate tasks in the trial plan.
The G1 family is relevant for trials involving humanoid movement; the manufacturer’s documentation distinguishes between models including the G1 and G1 EDU.
Booster Robotics presents the T1 as a humanoid platform for developers and documents API and development tools.
Fourier describes GR-2 and GRX as platforms for developing custom software and researching movement.
The Joybuy products below are intended for buying humanoid hardware for research and development trials; their product names alone do not prove that they are suitable for a specific industrial workplace.
No image available
The product title describes a humanoid robot with interactive features for trials involving movement and interaction.
It lists 29 Degrees of Freedom in total and 16 Joint Degrees of Freedom.
No image available
The product title describes the K1 Geek as a humanoid development platform for hardware trials.
The product title lists 48 TOPS AI and 22-DoF.
No image available
The product title describes the K1 Edu as a platform for research and development.
The product title lists 117 TOPS AI and 22-DoF; the description mentions software APIs and development tools.
No image available
The product name lists development access, movement control and autonomous navigation as features for development trials.
The product description recommends supervised operation and a distance of at least two metres. It also warns that the sensors may not detect every obstacle.
No image available
The product title mentions a camera and a language model for humanoid interaction trials.
Features listed include a Monocular Camera and 20 Degrees of Freedom.
Ask for details of the intended use, operating modes, range of movement, tools and moving objects, as well as the relevant manufacturer and integration documents. Employers must assess the risks associated with the actual use and identify suitable protective measures; the manufacturer’s assessment alone does not replace this step. This is explained by the DGUV Institute for Occupational Safety and Health and § 3 BetrSichV.
No, not as a blanket assurance. The stopping and monitoring functions, sensors and other measures required must be assessed in relation to the robot’s movements and proximity to employees in the configured workspace. DGUV guidance on collaborative robots is a useful reference, but it does not automatically approve a humanoid robot for use.
No. A simulation does not replace a risk assessment before real hardware is used. It can help you prepare scenarios and software workflows; for the actual workplace, you must assess the specific risks and define the protective measures required.
Ask for a quote covering the defined trial workflow, including commissioning, operation and development access, technical support, maintenance, spare parts, insurance and each party’s responsibilities. Also check which development rights are included with the rental.
No. Regulation (EU) 2023/1230 generally replaces the Machinery Directive from 20 January 2027, so that general date of application has not yet been reached in October 2026. Regardless, employers must assess the risks before work equipment is used. The timing and configuration are relevant to a specific delivery and commissioning; the European Union lists the date on which the Regulation applies.