Richtech Dex: mobile humanoid robot for industry
US company Richtech trains the AI of its robot Dex with real and simulated data to quickly make it fit for industrial use.
The industrial robot Dex has two powerful arms that, according to the manufacturer, can carry loads weighing 5 kilograms.
(Image: heise medien / André Kramer)
US-based robot systems manufacturer Richtech Robotics is presenting its mobile humanoid robot Dex at CES 2026 as a central component of its robotics platform for industrial and commercial applications.
The white robot measures 137 cm in its mobile version and 109 cm in its stationary version. It can carry a maximum of 5 kilograms. However, Dex is designed more for industrial and assembly tasks rather than as a load carrier. It resembles a kitchen appliance, classic white goods, more than an industrial robot. It doesn't look truly “humanoid.”
The manufacturer is showcasing it in conjunction with other robots, such as the coffee-making service robot Adam, the drink-mixing bar robot Scorpion, and the autonomous transport robot Titan, which can carry 200 kilograms or 590 kilograms depending on the version.
(Image: heise medien / André Kramer)
Head, Arms, and Wheels
The mobile version of Dex moves on wheels instead of legs. Richtech promises higher energy efficiency, lower maintenance, and stable locomotion in work environments shared with humans. According to the manufacturer, the chassis has a tight turning radius and can brake quickly.
The robot has two production arms, to which different grippers or tools can be attached at their ends. Four cameras assist with navigation, object recognition, and monitoring activities in changing environments.
Videos by heise
A system from Nvidia's Jetson Thor series serves as the computing unit. Because it processes sensor data locally, the robot is intended to be able to make quick decisions. In mobile operation, Dex achieves a runtime of around four hours per battery charge.
Software and AI Training
Richtech trains Dex using a combination of simulation and real-world operational data. To achieve this, the company uses open simulation environments such as Nvidia's Isaac Sim and Isaac Lab. In these virtual environments, the robot learns workflows that the manufacturer then transfers to real production environments. This is intended to allow the robot to be integrated quickly and safely into its work environment.
The control software is designed to learn new tasks through additional training data. Richtech draws on experience from more than 450 already installed robot systems. In parallel, the company is building its own database of real industrial applications in the USA, which it also intends to make available to third parties in the long term.
heise online is the official media partner of CES 2026
(akr)
