Fraunhofer IPA thus offers a neutral analysis platform for humanoid robots, which examines their performance, safety and industrial usability along six defined criteria. This is based on internationally recognized standards and industrial test procedures.

Between vision and reality: the missing standard in the hype market

Humanoid robots are currently very present: at trade fairs, in videos and in company presentations. However, there is often a clear gap between demonstration and real industrial use.

Where possible, the tests are based on established industry standards such as ISO 14644 for cleanrooms and ISO 10218 and ISO TS 15066 for functional safety in human-robot collaboration.

Six criteria for industrial suitability

The humanoid robots are evaluated in six core areas:

Technologies and basic skills:
Sensors, AI systems and gripping technologies are analyzed here. In addition, performance data such as speed, gripping force and payload are objectively measured using tracking and force sensors.

Complex skills:
Practical scenarios such as climbing stairs, overcoming obstacles and precision in movement and force dosage are tested. The tasks are deliberately challenging in order to keep future generations of robots comparable.

Cleanroom suitability:
Particle emissions according to ISO 14644-14, outgassing behavior and cleanability are evaluated - crucial for applications in the semiconductor, pharmaceutical or food industries.

Functional safety:
Stability, collision detection, force limitation and behavior in the event of system failures are examined. The collision tests are carried out using the same sensor technology as for collaborative industrial robots.

Cybersecurity:
Four test modules analyze vulnerability management, network architecture, system hardening and penetration resistance - an increasingly critical factor in networked production environments.

Energy efficiency:
Battery runtimes and energy consumption are measured in typical usage scenarios such as standing, walking or walking under load and on inclines.

First test with the Unitree G1: potential meets limits

The benchmark was applied comprehensively for the first time using the Unitree G1 as an example. The basis was a Unitree G1 EDU-4 with Dex3-1 3-finger hands and firmware version 1.04 delivered in May 2025.

The results show an ambivalent picture: on the one hand, the robot impressed with its stable self-balance and potential suitability for ISO class 5 cleanrooms. On the other hand, clear limitations also became apparent. For example, collisions resulted in forces of over 500 Newtons, well above the usual pain and safety limits.

In addition, the researchers discovered a critical Bluetooth security vulnerability in the tested software version, which would have allowed full remote control by attackers. This vulnerability has since been closed.

In terms of energy efficiency, the system achieved running times of 2 hours and 49 minutes in standby mode and 1 hour and 49 minutes in a combined walking and standing scenario.

Why comparability is becoming crucial

"Users can interpret the results directly and thus find the right humanoid for the right application," says Werner Kraus, Head of Research at Fraunhofer IPA.

The benchmark is intended to make humanoid robots systematically comparable with established automation solutions for the first time. This is particularly relevant against the backdrop of structural changes in the industry:

• Increasing pressure to automate due to demographic change
• High investment costs and therefore the need for a reliable basis for decision-making
• lack of binding safety standards for humanoids until at least 2028
• Increasing requirements for cyber security in networked production systems
• High sensitivity in cleanroom and hygiene areas