UBtech Walker X Humanoid Robot (Includes Lifting Frame)

The UBTECH Walker Tienkung DEX is a full-scale humanoid research platform for universities, research institutions, developers, and companies working on the next generation of intelligent robotics. The focus is on Physical AI, Embodied AI, humanoid locomotion, robot learning, and dexterous manipulation. Unlike a classic service robot, the Walker Tienkung DEX is consistently designed as an open development platform on which proprietary algorithms, movement strategies, and AI models can be tested under real-world conditions.

With a humanoid height of around 170 cm, 31 degrees of freedom excluding the dexterous hands, and AI computing power of up to 2070 TOPS, the robot forms a powerful foundation for demanding research projects. You can find further robotics solutions from the manufacturer on our overview page for UBTECH robots.

Full-size humanoid robot for realistic research

A decisive advantage of the Walker Tienkung DEX is its human-like structure. Instead of testing algorithms exclusively on small laboratory platforms or in simulations, motion planning and AI can be investigated on a full-scale humanoid robot.

The legs each have six degrees of freedom, and the arms each have seven. In addition, there is a movable waist, which enables complex full-body movements and coordinated movement sequences. The structure is therefore particularly suitable for research into humanoid locomotion, balance, full-body control, and human-robot interaction.

Those who wish to compare further platforms for research, industry, and humanoid automation can find them in our overview of humanoid robots.

Dexterous hands for grasping and manipulation

The "DEX" in Walker Tienkung DEX particularly stands for the expanded focus on dexterous manipulation. Multi-articulated robot hands enable the development of more complex grasping and interaction strategies than classic simple grippers.

This allows research teams, for example, to investigate how a humanoid robot recognizes objects, calculates suitable grip points, picks up items, and subsequently manipulates them in a controlled manner. This is relevant for pick & place, tool handling, sorting, assembly experiments, and training learning-based manipulation strategies, among other things.

The combination of two movable arms and dexterous hands makes the Walker Tienkung DEX particularly interesting for arm-hand coordination, robot learning, and vision-based manipulation. This also opens up numerous possibilities for industrial research projects revolving around flexible assembly and material handling. You can find suitable application examples in the Assembly & Handling section.

Up to 2070 TOPS for AI directly on the robot

Modern Embodied AI applications require considerable computing power. The Walker Tienkung DEX can be equipped with an NVIDIA Jetson AGX Thor platform and up to 2070 TOPS of AI computing power. This allows even more demanding perception, decision-making, and AI models to be processed directly on the robotics platform.

This is particularly relevant for Physical AI. While classic AI systems process information exclusively digitally, a humanoid robot must translate perception and decisions directly into physical actions. Camera and sensor data can be analyzed, for example, to derive movements, grasping strategies, or the next action from them.

As a result, the Walker Tienkung DEX is suitable as a research platform for multimodal AI, vision-language models, autonomous decision-making processes, and new forms of human-robot interaction.

Open platform with ROS 2, SDK, and URDF

One of the most important features of the Walker Tienkung DEX is its open development environment. The platform supports ROS 2 and provides open interfaces for joints and sensors. In addition, development resources, example code, and high-resolution URDF models are available.

This allows developers to implement their own controllers, motion algorithms, perception systems, and AI workflows. The complete URDF model contains kinematic and dynamic descriptions, visual models, as well as collision meshes, and can be used for visualization and simulation, among other things.

For universities and research institutions in particular, an open platform significantly reduces the effort required because not every basic interface needs to be developed from scratch.

Simulation, reinforcement learning, and sim-to-real

Before new movements or AI algorithms are executed on a real humanoid robot, they can first be tested in a simulation environment. For this purpose, UBTECH provides, among other things, URDF models and open training frameworks.

Research approaches revolving around reinforcement learning, embodied manipulation, and sim-to-real transfer are supported. Movement strategies can be trained virtually and subsequently transferred to the real Walker Tienkung DEX and further optimized.

This workflow is particularly relevant for modern robotics research, as large amounts of training data can be generated without having to carry out every movement directly on the physical robot.

Teleoperation and data collection for robot learning

Teleoperation and data collection also play an important role in the development of humanoid robotics. The Walker Tienkung DEX supports corresponding development workflows so that human movement and demonstration data can be captured for research projects.

Such data form a foundation for imitation learning and other learning-based methods. A human can, for example, demonstrate how a specific object is grasped or a task is executed. These demonstrations can subsequently be used as training data for AI models.

UBTECH also describes datasets with RGB-D recordings, robot states, end-effector data, and natural language task descriptions as part of its open Embodied AI orientation.

Research platform for universities and development teams

The Walker Tienkung DEX is aimed particularly at universities, research institutes, robotics labs, and R&D corporate departments. Topics such as motion control, artificial intelligence, sensor fusion, robot learning, grasp planning, and autonomous robotics can be investigated on a shared full-scale platform.

The robot is thus suitable for both scientific research projects and practice-oriented courses and development projects. You can find further systems specifically for teaching, universities, and robotics education among our educational robots.

Using the UBTECH Walker Tienkung DEX for Physical AI projects

The UBTECH Walker Tienkung DEX combines full-body humanoid movement, powerful AI hardware, dexterous manipulation, and an open development architecture. This creates a platform on which new approaches to Physical AI can not only be simulated but also investigated on a real humanoid system.

Whether reinforcement learning, sim-to-real, teleoperation, autonomous manipulation, humanoid locomotion, or multimodal AI: which configuration makes sense depends on the respective research goal. Terra Robotics provides support with the selection, configuration, and integration of the Walker Tienkung DEX and can, together with universities, research institutions, and companies, examine which hardware, sensor technology, and development options are required for the planned project.