NineRay Technology unveiled its RayNex G3 humanoid robot in Hong Kong on August 17, 2026, marking a significant step toward bringing embodied AI systems into real-world commercial applications. The 173-centimeter-tall robot, weighing less than 60 kilograms, can carry 30 kilograms during operation, deadlift 100 to 200 kilograms, and hold 60 kilograms with both arms, making it capable of handling both heavy logistics tasks and precision work.

What Makes RayNex G3 Different From Other Humanoid Robots?

The RayNex G3 features 32 degrees of freedom (DOF), a measure of how many independent ways a system or its parts can move. The robot has seven degrees of freedom in each arm, six in each leg, and three each in the neck and waist. This level of articulation allows the robot to perform both bulk conveyance tasks in warehouses and intricate precision maneuvers that require fine motor control.

Embodied AI refers to artificial intelligence systems housed within physical robots and mechanical bodies, enabling them to perceive their environment, learn from events as they unfold, and carry out physical operations. This differs from traditional AI systems that exist only in software form.

During the launch event at Hong Kong Science Park, the robot demonstrated its capabilities by striding across the stage while carrying over 60 kilograms, then returning for another demonstration while walking with a barbell to simulate a heavy side-lift.

How Is Hong Kong Positioning Itself in the Global Robotics Market?

NineRay founder Xu Zhigen stated that the company has positioned Hong Kong as its "base for globalization," emphasizing the city's advantages for robotics development and commercialization. According to Xu, Hong Kong's global market access, deep pools of talent and capital, and robust research expertise from world-leading universities have been instrumental to the company's growth.

Xu Zhigen

Terry Wong Ping-sau, CEO of Hong Kong Science and Technology Parks Corporation, expressed optimism about NineRay's presence in the city. Wong noted that he hoped NineRay would continue to leverage the Guangdong-Hong Kong-Macao Greater Bay Area's full industrial ecosystem to refine core technologies, explore real-world applications, and move humanoid robots from laboratory settings into commercial deployment at scale.

The Hong Kong Science Park currently accommodates over 2,500 technology firms, including NineRay's regional headquarters, making it one of the city's most active innovation hubs.

What Real-World Applications Are Already in Development?

Xu disclosed that multiple firms spanning diverse industries have already approached NineRay's development team with proposed applications for the RayNex G3. These industries include:

• Food and Beverage Production: Companies in this sector are exploring how the robot can handle repetitive assembly, packaging, and material handling tasks.
• Warehouse and Logistics Operations: The robot's ability to carry heavy loads and navigate standard indoor and outdoor work settings makes it suitable for distribution centers.
• Nuclear Power Sector: Organizations in this industry are investigating applications where the robot could perform tasks in environments that may be hazardous to human workers.

How Does Touch Sensing Enhance Physical AI Capabilities?

Beyond mechanical design, the field of embodied AI is advancing through tactile sensing technology. Touch data collected from robot surfaces during interaction with objects, people, tools, or environments complements visual perception and enables more sophisticated manipulation and safety features.

Tactile sensing becomes especially valuable when a robot's hand or gripper blocks the camera's view of the work area. In these situations, touch data provides critical feedback about slip, seating, deformation, force patterns, and contact timing that vision alone cannot capture.

A complete tactile data workflow for embodied AI systems should include capture, alignment, storage, and interpretation stages. This means preserving each contact event with its timestamp, body frame reference, calibrated sensor value, and corresponding robot action so that the signal can be replayed, compared, and used for training and evaluation outside the original demonstration.

Steps to Building Effective Tactile Sensing Into Robots

• Capture Phase: Record the contact event, timestamp, body frame location, calibrated sensor value, and robot action simultaneously to create a complete record of the interaction.
• Alignment Phase: Synchronize tactile data with joint state, vision data, control commands, and task phase information so all sensor streams are temporally aligned.
• Storage Phase: Retain calibration metadata, sensor location, measurement units, sampling rate, and any failure notes to ensure the data remains interpretable by other teams and systems.
• Use Phase: Expose tactile features for grasp control, safety checks, replay diagnostics, or machine learning systems that can learn from the contact patterns.

Physical AI systems combine multiple complementary sensing channels rather than relying on a single sensor type. Vision describes visible geometry and motion, language encodes task goals and semantic context, proprioception reports the robot's body state and configuration, and touch measures the contact that results when an action meets the physical world.

Lu Yang, associate dean overseeing mainland affairs at the University of Hong Kong's Faculty of Engineering, observed that humanoid robots represent one of the primary physical vessels for embodied intelligence. He noted that Hong Kong is already home to a substantial community of embodied intelligence specialists, and that the city's universities are now building on this foundation to attract more global talent for robotics research and industrial application development.

"It's a pivotal direction, not only for fostering new quality productive forces and powering new industrialization, but also for propelling AI-empowered innovation forward," stated Lu Yang.

Lu Yang, Associate Dean, University of Hong Kong Faculty of Engineering

The convergence of advanced mechanical design, tactile sensing, and AI integration suggests that humanoid robots are transitioning from research prototypes to practical tools for industry. Hong Kong's emergence as a development and commercialization hub reflects broader momentum in the embodied AI field, where multiple companies and research institutions are working to solve the technical and logistical challenges of deploying physical AI systems at scale.