A robot learns to handle bulky objects like humans do after just one lesson

September 3, 2025 report

The GIST A robot learns to handle bulky objects like humans do after just one lesson

Paul Arnold

contributing writer

Gaby Clark

scientific editor

Robert Egan

associate editor

Editors' notes

This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

peer-reviewed publication

trusted source

proofread

A robot learns to handle bulky objects like humans do after just one lesson
Pictures showing the (A) rigid air-filled chamber stand-ins and paws used for compliance experiments. (B) close-up picture of the rigid air-filled chamber stand-ins showing the cutouts added to avoid reducing the range of motion due to self-collisions compared to the soft air-filled chambers. Credit: Science Robotics (2025). DOI: 10.1126/scirobotics.ads6790

For all their technological brilliance, from navigating distant planets to performing complex surgery, robots still struggle with a few basic human tasks. One of the most significant challenges is dexterity, which refers to the ability to grasp, hold and manipulate objects. Until now, that is. Scientists from the Toyota Research Institute in Massachusetts have trained a robot to use its entire body to handle large objects, much like humans do.

We use a variety of skills to pick up and manage objects, from small, precise hand movements to large-muscle movements involving the arms, legs and trunk (gross motor skills). Robots find these larger movements difficult, such as lifting and holding onto a large box, because they require constant, complex adjustments to prevent the object from fumbling or dropping.

In a study published in the journal Science Robotics, the research team demonstrated how a humanoid upper-body robot, named Punyo, was able to move a large water jug onto its shoulder. It could also pick up and hold onto a large box. To guide its movements, the robot relied on feedback from its soft, pressure-sensing skin and joint sensors.

Hardware rollout of the blind (no object pose tracking, only proprioceptive-and-tactile 1072 feedback), jug-over-shoulder policy. Credit: Jose A. Barreiros

The key to the robot's success was the softness of its body (passive compliance) and the flexibility programmed into its joints (active compliance). When compared to a rigid version of the robot, it performed much better. "Adding any type of compliance (that is, passive, active, or both) is beneficial, showing an increase of 206% in the success rate on average compared to no compliance," the scientists wrote.

What's more, the robot required minimal learning. The researchers taught Punyo using a method called example-guided reinforcement learning. They showed it what to do with just one demonstration in a virtual environment, and then it practiced on its own until it figured out how to do it. "A single teleoperated demonstration collected in simulation is sufficient to train policies for stylized, contact-rich motions," the research team added.

Hardware rollout of the cardboard-box-pivot-and-lift policy. Credit: Jose A. Barreiros

More useful robots

This technology is a significant step toward creating robots that can be even more helpful in our daily lives. For example, they could safely and reliably move large, unwieldy objects, such as furniture in the home or heavy packages in a warehouse. They could even help with tasks in a care setting, such as assisting people with mobility challenges. And the robots wouldn't need to be meticulously programmed. Instead, they would be able to learn human-like skills from just a single demonstration.

Written for you by our author Paul Arnold, edited by Gaby Clark, and fact-checked and reviewed by Robert Egan—this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive. If this reporting matters to you, please consider a donation (especially monthly). You'll get an ad-free account as a thank-you.

More information: Jose A. Barreiros et al, Learning contact-rich whole-body manipulation with example-guided reinforcement learning, Science Robotics (2025). DOI: 10.1126/scirobotics.ads6790

Journal information: Science Robotics

© 2025 Science X Network

Citation: A robot learns to handle bulky objects like humans do after just one lesson (2025, September 3) retrieved 3 September 2025 from https://techxplore.com/news/2025-09-robot-bulky-humans-lesson.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

Explore further

Robotic hand uses compliant materials to grasp objects with human-like dexterity 0 shares

Feedback to editors