Artificial Intelligence 10 min read

How Tencent’s TRX‑Hand and TRX‑Arm Achieve Human‑like Dexterous Manipulation

On April 25, Tencent’s Robotics X lab unveiled its self‑developed TRX‑Hand and TRX‑Arm, a three‑finger dexterous hand and a seven‑degree‑of‑freedom arm that combine advanced hybrid actuation, high‑resolution tactile sensors, multi‑modal perception and real‑time planning to perform complex tasks such as cocktail mixing, object catching, and precise insertion.

Tencent Tech
Tencent Tech
Tencent Tech
How Tencent’s TRX‑Hand and TRX‑Arm Achieve Human‑like Dexterous Manipulation

Introducing TRX‑Hand and TRX‑Arm

Tencent Robotics X laboratory announced the first demonstration of its self‑developed dexterous hand TRX‑Hand and high‑performance seven‑degree‑of‑freedom robotic arm TRX‑Arm . The hand features three fingers with eight independently controllable joints, weighs only 1.16 kg, and can exert up to 15 N fingertip force while moving at speeds above 600°/s.

TRX‑Hand demonstration
TRX‑Hand demonstration

Mechanical Design and Actuation

The hand uses a patented rigid‑soft hybrid drive and a self‑designed high‑power‑density actuator, providing both high agility and high load capacity. Its lightweight structure and fast joint speeds enable rapid grasping, dynamic throwing, and precise manipulation of objects of various shapes and sizes.

Multi‑modal Sensing

TRX‑Hand integrates high‑sensitivity flexible tactile sensor arrays on the fingertips, finger pads, and palm, a miniature LiDAR and proximity sensor on the palm, and angle sensors on each joint. This dense sensing suite allows accurate perception of the hand’s state and the objects it interacts with.

TRX‑Hand tactile sensors
TRX‑Hand tactile sensors

Perception and Control Algorithms

By fusing visual, proximity, tactile, and force data, the system builds a comprehensive model of the environment. Multi‑sensor fusion enables robust object recognition and pose estimation even for transparent, reflective, or occluded items. Real‑time planning algorithms compute optimal arm‑hand configurations and safe trajectories, while adaptive control adjusts motions based on sensor feedback.

Demonstrations and Tasks

The platform performed a series of complex tasks, including cocktail mixing that required over 30 sequential operations such as grasping, pouring, rotating, shaking, and dynamic hand‑hand coordination. It also demonstrated high‑speed object catching, precise insertion of thin, reflective stirrers into narrow holes, and in‑hand manipulation of objects using deep reinforcement learning and sim‑to‑real techniques.

Robot performing cocktail mixing
Robot performing cocktail mixing

TRX‑Arm Specifications

The TRX‑Arm employs a rope‑driven differential drive that reduces friction and inertia, enabling high‑dynamic motion with a maximum speed of 7.4 m/s, acceleration of 44.5 m/s², and payload capacity over 6 kg. Its forearm houses a 768‑point tactile array refreshed at 1000 Hz, supporting precise real‑time positioning of objects such as bottles during rolling and balancing tasks.

TRX‑Arm demonstration
TRX‑Arm demonstration

Future Directions

Tencent Robotics X plans to further integrate TRX‑Hand and TRX‑Arm with advanced deep‑learning algorithms to enhance dexterous manipulation and solve more complex real‑world problems, targeting applications in eldercare, home services, and intelligent manufacturing.

aiRoboticssensor fusiondexterous manipulationrobotic hand
Tencent Tech
Written by

Tencent Tech

Tencent's official tech account. Delivering quality technical content to serve developers.

0 followers
Reader feedback

How this landed with the community

login Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.