World’s first fully electric robot boasts 550 trillion ops

Interesting Engineering carries an article about tiangong which combines visual perception sensors, 3D vision sensors, and high-precision IMU to tackle motion control challenges effectively, Caliber.Az reprints the article.

The humanoid robot domain has seen intense activity in recent times with multiple startups vying to get their prototypes into the market.

Now, a state-backed robotics innovation center in Beijing has launched its first electric humanoid “Tiangong”.

The robot was revealed at the Beijing Economic-Technological Development Area on May 4, which coincided with the announcement of numerous AI technological advancements at the 2024 Zhongguancun Forum (ZGC Forum).

Tiangong, crafted by the Beijing Humanoid Robot Innovation Center Company, the general-purpose humanoid boasts the capability to sustain a consistent pace of 3.73 miles per hour (6 kilometers per hour).

The Chinese government plans to facilitate widespread access to a humanoid robot platform through Tiangong, largely offering it as an open-source solution.

Agile and adaptive humanoid

Tiangong’s head, arms, and legs can all be moved in multiple directions thanks to its entirely electric actuators.

The 163 cm tall Tiangong is outfitted with several sensors for visual perception and has a processing speed of 550 trillion operations per second. It also features high-precision six-axis force sensors to provide accurate force feedback.

In addition to its fully electric operation, the humanoid robot boasts a lightweight construction (94.7 pounds or 43 kilograms), ensuring stability during movement, particularly while running. Its natural running motion is achieved through State Memory-based Predictive Reinforcement Imitation Learning, a cutting-edge skill in motion learning utilized by humanoid robots.

Furthermore, the Tiangong humanoid robot integrates a variety of visual perception sensors, 3D vision sensors, and a high-precision inertial measurements unit (IMU). This comprehensive sensor suite collectively addresses the robot’s fundamental motion control challenges.

As seen in a video shared, noteworthy is Tiangong’s ability to navigate slopes and stairs seamlessly, even without visual input. Moreover, it demonstrated enhanced adaptability by swiftly modifying its gaits to recover from stumbles or falls.

According to its developers, Tiangong has already shown human-like running skills and provides open-source compatibility for future development, enabling wider commercial applications.

China’s ambitious push

China has laid out a detailed plan to help itself take the lead in the domain of humanoid robot development.

A vision strategy was announced in November 2023 by the Chinese Ministry of Industry and Information Technology aims to begin mass-producing humanoid robots in 2025. According to it, by 2027, these robots will also be able to think, learn, and create.

As a part of it, the Economic and Technological Development Area, Beijing’s center for the robotics industry, has drawn together 110 robotics ecosystem businesses. They aim to establish an entire industrial chain system that includes apps, whole machines, and key components.

The center is focused on developing parent platforms for both software and hardware. It also aims to solve critical issues facing the industry, like open-source operating systems and operation control systems, which require immediate attention.

In line with the government’s strategy, Tiangong’s open-source design distinguishes it from commercial robotics companies, enabling academic institutions and robotics firms to utilize it as a foundation for innovation.

China’s humanoid robot domain is flourishing, featuring advanced models from various private enterprises. Notable examples include UBtech Robotics’ Walker S, LimX Dynamics’ CL-1, Fourier Intelligence’s GR-1, Astribot’s S1, and Unitree’s H1.

Add to it, experts suggest that China’s considerable progress in integrating AI capabilities into daily life could accelerate the advancement of its humanoid robots.