Researchers from the University of Bristol have achieved a significant breakthrough in robotics, developing a two-armed robot with tactile sensitivity comparable to humans. This groundbreaking achievement has opened up new possibilities in industries such as agriculture, household services, and medical prosthetics.
The unique system, known as Bi-Touch, utilizes artificial intelligence to analyze data gathered from tactile and proprioceptive sensors. Through the interaction with these sensors, the robot can process information about an object’s texture, temperature, shape, as well as the position and movement of its hands.
The authors of the research state that the Bi-Touch system allows for training in a simulator in just a few hours. This means that AI agents can be deployed into real-world settings without the need for additional training.
The researchers employed a deep learning with reinforcement (Deep-RL) approach to teach the robot. By experimenting with various actions and analyzing their outcomes, the robot learns how to interact with objects. Successful completion of tasks earns the robot rewards, which incentivize it to further improve its performance. The robot has already demonstrated a high level of “dexterity” by delicately handling fragile objects, such as chips.
The robot carefully moves the chips without breaking them.
In the field of robotics, “dexterity” refers to the robot’s ability to manipulate objects effectively. This can include lifting, moving, or delicately interacting with fragile or complex materials.
This development will serve as the foundation for future research as the system’s code will be made available to the scientific community. According to the researchers, the Bi-Touch system has the potential to revolutionize not only the industrial sector but also the field of medicine, particularly in the creation of highly sensitive artificial limbs. Furthermore, this robot’s level of sensitivity will enable the acceleration of automation in various processes that were previously impossible.