When choosing a humanoid robot for research or teaching, start by identifying the control, sensor and data access your specific project requires. Compact development platforms may suit teaching and AI prototyping. Larger research platforms can be a good fit for experiments involving dynamic locomotion or torque control, provided your lab has the space, safety measures and integration capabilities they require. For every decision, documentation for a model family is no substitute for confirming the exact variant you are ordering.
To develop their own control software, these teams need documented interfaces, suitable simulation models and access to the state and sensor data they require. The key question is whether these capabilities are available on the configuration being supplied.
They need to combine movement and state data with a test space and safety measures that allow a bipedal system to be tested repeatedly. If experiments take place near people, a project-specific risk assessment must be part of the plan.
For grasping and interaction experiments, the hands, arms and sensors actually supplied are what matter. If camera, audio or interaction data involving people are collected, how that data will be processed must also be agreed before testing begins.
Start with the experiment you plan to run, then check the exact variant, software access and requirements for operating it in your lab.
| Requirement | What to check before buying | Who it matters most to |
|---|---|---|
| Development access | Confirm in writing which SDK and API features are available for the variant being ordered, along with the ROS 2 interfaces, sensor data and simulation models you need. | If I want my own algorithms to run on the robot, I need access to the specific control and data layers my project requires. This is the main priority for around 36% of consumers in university and research teams. |
| Safe lab operation | Establish the movement area, safety measures and a project-specific risk assessment before conducting experiments with people. | When carrying out repeated movement trials, I need to account for falls and potential collisions as part of the lab procedure. Around 29% of consumers give safe testing the highest priority. |
| Manipulation and sensors | Check the hand and arm configuration, along with the camera, depth or audio data required for the experiment, against the exact variant. | For grasping or interaction research, a programmable robot is only useful to me if its specific equipment suits the task. The right manipulation and sensor configuration is the main concern for around 21% of consumers. |
| Service and integration | Agree on what is included, maintenance, spare parts, training and responsibilities for network integration. | If the platform is part of a teaching or project plan, I need clear responsibilities for commissioning and ongoing support. For around 14% of consumers, these organisational requirements come first. |
Unitree describes different configurations for the G1 EDU, while Booster Robotics also offers T1 variants with different numbers of degrees of freedom. A shared model name does not tell you exactly which hands or research features are included.
Ask the supplier to confirm the variant identifier and what is included, including the hand type, sensors, computing hardware and software access.
Unitree’s SDK documentation for G1 developer configurations does not prove that every G1 offer provides the same access. The Unitree G1 listing on Joybuy explicitly states that the SDK and API are currently unavailable.
Before building lab software around the robot, check the variant identifier on offer against the documentation and confirm the promised interfaces in writing.
Movement and collision risks also depend on the setup and how the experiment is run. The German Social Accident Insurance (DGUV) recommends risk assessments and suitable impact force measurements for collaborative robot scenarios.
Before approving the experiment, plan the movement area, access, transport and emergency procedures with your lab manager and health and safety team.
You may need to allow additional preparation time for safety zones, measurements and training.
Camera, audio and interaction data can involve people. The European Data Protection Board stresses the need to protect data subjects in scientific research, while the University of Hamburg outlines how to document the processing of personal data in research labs.
Before collecting data, agree the purpose, access, storage location, deletion process and potential recipients with your data protection and IT security teams.
The G1 EDU and H1 are suited to different research needs: more compact development projects and experiments using a larger platform, respectively.
According to Booster Robotics, the T1 range is designed for development, teaching and research projects using custom software.
TALOS is a research platform for work on whole-body control, locomotion and manipulation.
These two humanoid platforms may suit teaching and development projects. Before choosing, check in particular which software and sensor data access is available for the specific offer. Documentation for another model variant is no substitute.
No image available
The X2 Ultra may suit projects using custom control software, provided the development access you need is confirmed in writing for the variant on offer.
For perception or manipulation experiments, teams should also check the accessible sensor data, usage rights and limits on the planned arm movements against the offer.
No image available
The K1 Edu may be a suitable platform for teaching and AI prototyping, provided its actual API features meet the needs of your experiment.
The listing specifies 22 degrees of freedom and 117 TOPS of computing power. Check the sensor access, software permissions and lab operating requirements for the variant on offer.
No. On its own, this doesn’t tell you which features are accessible on the variant you are ordering. Ask for specific confirmation of the control layer, sensor data, ROS 2 integration, simulation model and usage rights. Unitree’s G1 developer documentation does not prove that every G1 offer provides the same access.
If personal data is involved, clarify the research purpose and how the data will be processed before collecting it. The European Data Protection Board stresses the importance of protecting individuals in scientific research. The University of Hamburg also outlines the documentation of purposes, data categories, recipients and retention periods. Include your data protection and IT security teams in the experimental design.
No. Suitability depends on the specific system and experimental setup. For collaborative robotics, the German Social Accident Insurance (DGUV) describes the need for a risk assessment and refers to suitable impact force measurements. You should also take movement, tools, space and safety measures into account.
Request documents for the exact model variant, particularly details of software access, operating and safety measures, what is included, network connections and service arrangements. Also check which conformity documents apply to the specific system and its intended use. These do not replace a risk assessment for the lab experiment.