Home AI News Robotic Hand Masters In-Hand Rotation Solely Through Touch, No Vision Needed

Robotic Hand Masters In-Hand Rotation Solely Through Touch, No Vision Needed

Robotic Hand Masters In-Hand Rotation Solely Through Touch, No Vision Needed

Robotic Hand Can Rotate Objects Using Touch Alone, UC San Diego Engineers Find

A team of engineers at the University of California San Diego has developed a groundbreaking approach that allows a robotic hand to rotate objects solely through touch, without relying on vision. Inspired by the way humans effortlessly handle objects without seeing them, this new technique has the potential to revolutionize robotic manipulation and enable robots to perform tasks in the dark.

How it Works

The researchers equipped a four-fingered robotic hand with 16 touch sensors, which were attached to the palm and fingers. These touch sensors, costing only $12 each, have a simple function: to detect if an object is touching them or not. Unlike previous approaches that use a few expensive high-resolution touch sensors, this method utilizes many low-cost, low-resolution touch sensors spread over a larger area of the hand.

This unique approach has several advantages. Firstly, having a large number of sensors increases the chances of contact with the object, enhancing the system’s sensing ability. Secondly, the binary signals provided by these sensors (touch or no touch) are much easier to simulate and transfer to the real world, compared to high-resolution touch sensors that provide texture information. Lastly, and importantly, this method does not rely on vision, unlike other approaches.

The Successful Experiment

To train the robotic hand, the researchers ran simulations of a virtual hand rotating objects with different shapes. The system determined which sensors were being touched and the positions of the hand’s joints during each rotation. Using this information, the system instructed the hand on the next actions it needed to take.

The team then put their system to the test by using the real-life robotic hand on various objects it had never encountered before, such as a tomato, pepper, can of peanut butter, and a toy rubber duck. Surprisingly, the hand was able to rotate all the objects without stalling or losing its grip. Although more complex objects took longer to rotate, the system successfully rotated them around different axes.

Potential Future Applications

While this study focuses on the rotation of objects using touch alone, the team aims to expand the capabilities of robotic hands in the future. They are currently working on techniques to enable robotic hands to catch, throw, and juggle objects. Mastering in-hand manipulation is a complex skill for robots, but if achieved, it could greatly expand the range of tasks they can perform.

This breakthrough research, titled “Rotating without Seeing: Towards In-hand Dexterity through Touch,” was presented at the 2023 Robotics: Science and Systems Conference. The paper’s co-authors include Binghao Huang, Yuzhe Qin, Xiaolong Wang, Zhao-Heng Yin, and Qifeng Chen.

Source link


Please enter your comment!
Please enter your name here