Reward AI Demonstrates Humanoid Robot Delivery Skills
The U.S. robotics startup uses sensor gloves to train its new OM-1 humanoid model for common household and delivery tasks.
Updated on Oct. 31, 2026 in Robotics
Live Poll
Do you believe humanoid robots will soon reliably handle complex tasks in everyday settings?
Reward AI has released a demo video showcasing its OM-1 humanoid robot performing tasks like mopping and moving parcels. The company trains the robot by recording human hand motions and gripping force through specialized sensor gloves.
Why it matters
The company relies on sensor gloves to generate training data, a strategy designed to bypass the industry-wide shortage of high-quality robot training material. This approach offers an alternative to mass simulation methods used by other firms to accelerate development.
Reward AI utilizes its OM-1 model to handle manual tasks through human-captured motion data. Comparatively, Nvidia utilizes 4,096 simultaneous virtual environments to train its MicroDuck robotics platforms.
The players
Reward AI
This U.S.-based robotics startup focuses on developing AI models for humanoid robot navigation and manipulation.
A global technology company that develops AI-driven hardware, including its Gemini Robotics line.
Nvidia
A computer hardware and software company that provides the virtual simulation infrastructure used for training advanced robotics.
Wang Xinxing
The founder of the robotics firm Unitree who provides expert projections on the future capabilities of industrial robots.
The details
By using sensor gloves to replicate human motor skills, Reward AI aims to imbue the OM-1 robot with the dexterity needed for delivery work. This method contrasts with virtual training techniques, such as those used by Nvidia, which focus on amassing experience in simulated training grounds to reduce physical wear.
Timeline
July 2026: Google unveiled the Gemini Robotics 2 humanoid.
October 23, 2026: Nvidia released a demonstration video of its MicroDuck robots.
October 27, 2026: Reward AI released its humanoid robot demo video.
The Tech Race
The development of the OM-1 robot aligns with a broader industry goal to reach an 80 percent task performance threshold across most new environments. This evolution marks a transition from simple automation toward robots capable of managing varied physical tasks in unstructured spaces.
As these robots evolve toward handling household tasks, users may eventually see improvements in automated delivery and domestic chore assistance. However, adoption remains dependent on the robots achieving consistent reliability in diverse, real-world environments.
The takeaway
The move toward human-motion training for robots represents an effort to bridge the data gap that currently hinders widespread adoption. Developing systems that can learn directly from human experience could eventually make robot labor more viable for everyday logistics.
Further reading
For more information on the current state of autonomous machines, see our Robotics coverage.
Source note: This article includes information reported by 경향신문.
Live Poll
Do you believe humanoid robots will soon reliably handle complex tasks in everyday settings?
