Researchers from the University of Cambridge have developed a robotic sensor that utilizes artificial intelligence (AI) techniques to read braille at speeds double that of most human readers. Using machine learning algorithms, the team trained the robot to quickly glide over lines of braille text. The results were impressive, with the robot achieving a reading speed of 315 words per minute with close to 90% accuracy. Although not intended as an assistive technology, the researchers believe that the high sensitivity required to read braille makes it an ideal test for the development of robot hands or prosthetics with similar sensitivity to human fingertips.
Human fingertips possess remarkable sensitivity, allowing us to gather information about the world through touch. They can detect subtle changes in texture or gauge the amount of force needed to hold or manipulate an object. Replicating this level of sensitivity in a robotic hand, while ensuring energy efficiency, presents a significant engineering challenge. Led by Professor Fumiya Iida, the researchers in Cambridge are working on solutions to impart skills that come naturally to humans but prove difficult for robots.
Parth Potdar, an undergraduate at Pembroke College and one of the researchers involved, explained, The softness of human fingertips is one of the reasons we’re able to grip things with the right amount of pressure. For robotics, softness is a useful characteristic, but you also need lots of sensor information, and it’s tricky to have both at once, especially when dealing with flexible or deformable surfaces.
Braille serves as an ideal test for a robot ‘fingertip’ due to its high sensitivity requirements, given the close proximity of the dots in each letter pattern. By using an off-the-shelf sensor, the researchers were able to develop a robotic braille reader that closely mimicked human reading behavior.
While existing robotic braille readers work in a static way, touching one letter pattern at a time, our robot reads braille more realistically and efficiently, said co-author David Hardman. The robot utilized a camera in its fingertip, combining the information from the camera and the sensor. However, processing the images to remove motion blur was a challenging task that incurred time and energy costs.
To overcome this obstacle, the team developed machine learning algorithms that enabled the robot to deblur the images and recognize the letters. The algorithms were trained using sharp images of braille with artificial blur added. After successfully training the algorithm, a computer vision model was used to detect and classify each braille character.
Once the algorithms were incorporated into the system, the researchers conducted tests where the robotic reader quickly slid along rows of braille characters. The results were remarkable, with the robot achieving a reading speed of 315 words per minute with an accuracy rate of 87%. This makes the robotic braille reader twice as fast as a human braille reader, achieving a similar level of accuracy.
Braille reading speed is a great way to measure the dynamic performance of tactile sensing systems, so our findings could be applicable beyond braille, for applications like detecting surface textures or slippage in robotic manipulation, added Potdar.
Looking ahead, the researchers envision scaling the technology to the size of a humanoid hand or even applying it to robotic skin. The project received partial support from the Samsung Global Research Outreach Program.
The development of this robotic sensor marks a significant advancement in the field of robotics, opening up possibilities for improved automation and assistive technologies. By gaining insights from human touch sensitivity, researchers are bringing us closer to a future where robotic hands and prosthetics can fully replicate the intricate abilities of the human touch.
