Research Platforms

Buying advice

Read the warranty with your experiments in mind

Research use is hard on a robot, and a warranty often assumes the machine will be used as shipped. Ask each maker whether opening the casing, adding sensors or flashing your own firmware voids cover, and whether repairs can be done in your own workshop. By October 2026, listed prices here spanned roughly $1,370 (Noetix Bumi) to $60,000 (Xiaomi CyberOne), and no listing was quote-only. Booster Robotics appears twice, with the T1 and the K1 Geek, so ask whether code and tools carry over between them before you mix both in one course. Check your institution's procurement rules early, since some grants require competing quotes.

Make your locomotion results comparable

Published gait results are hard to compare when every lab builds its own course. ISO 18646-5:2026 sets out test methods for evaluating how legged robots move, and it leaves safety verification out of scope, so treat it as a performance yardstick rather than a safety certificate. Ask each maker whether it has measured its robot against these methods, and consider laying out your own test area to match them so results survive a change of platform. Record firmware and controller versions alongside every run, because a maker's update can shift behaviour between experiments. If the maker allows it, keep a study robot off automatic updates until the work is written up.

Plan the data before recording begins

Humanoid research increasingly relies on recorded demonstrations and sensor logs, and those files outlive any single robot. ISO/CD 26264-1, a draft standard on humanoid robot datasets, covers the whole life of a dataset from planning to decommissioning, including quality, security and privacy. Use it as a checklist when you write your data management plan and ethics application, especially if cameras will capture students or members of the public. Ask makers whether raw sensor streams are available without a cloud account, which formats they export, and whether any telemetry is sent back to them by default.

What to look for

• SDK access, and support for the simulators and frameworks your lab already uses
• Degrees of freedom in legs, arms and hands for the research you have planned
• Repairability, spare parts and how well the robot copes with falls during testing

Common questions

Which research humanoid suits a teaching lab?

Smaller, lower-cost platforms are often easier to run with students, while full-size robots suit locomotion and manipulation research. Compare height, weight and SDK access before you choose.

Do research platforms come with open software?

Openness varies by maker and sometimes by edition. Check each listing for SDK and low-level control access, and confirm it in your quote request if your work depends on it.

Sources and further reading

• ISO 18646-5:2026 Robotics — Performance criteria and related test methods for service robots — Part 5: Locomotion for legged robots — International Organization for Standardization.
  Test methods for measuring how legged robots walk, so humanoid platforms can be compared on equal terms.
• ISO/CD 26264-1 Humanoid robot datasets — Part 1: General requirements — International Organization for Standardization.
  Draft standard on planning, quality, security and privacy of the datasets used to train humanoid robots.
• Humanoid Robots: Vision and Reality (IFR position paper) — International Federation of Robotics.
  IFR's view of where humanoid robots stand, the technology behind them and the sectors they are aimed at.