Researchers at the University of Cambridge have made a significant breakthrough in robotic technology by creating a robotic sensor that uses artificial intelligence to read braille. This innovation allows the robot to interpret braille text at a speed of 315 words per minute with around 90% accuracy, which is nearly twice the rate of an average human reader.
This project, detailed in the journal IEEE Robotics and Automation Letters, was not primarily aimed at developing assistive technology. Instead, its focus was on advancing robotic sensitivity, potentially benefiting the creation of robot hands or prosthetics with a sensitivity similar to human fingertips.
Human fingertips are incredibly sensitive, aiding us in detecting subtle changes in textures or applying the correct amount of force for tasks like picking up an egg or a bowling ball. Mimicking this sensitivity in robots, while maintaining energy efficiency, poses a substantial challenge for engineers.
At the Department of Engineering in Cambridge, under Professor Fumiya Iida, the research team is tackling this challenge. Parth Potdar, a Cambridge Engineering student and the paper’s primary author, emphasizes the importance of softness in human fingertips for effective gripping. Achieving a balance between this softness and the need for extensive sensor information, especially on flexible surfaces, is a key hurdle in robotics.
Braille reading is an excellent test for a robot ‘fingertip’ due to the high sensitivity required to discern closely spaced dots. The team utilized a commercially available sensor to create a braille reader that more closely resembles human reading patterns.
David Hardman, another co-author from the Engineering Department, notes that existing robotic braille readers are less efficient as they read one letter at a time, unlike humans. The team’s goal was to develop a more realistic and efficient approach.
The robotic sensor includes a camera at its ‘fingertip’, combining camera data and sensor input for reading. Potdar points out the complexity of this task, particularly in processing images to remove motion blur, a process that is both time and energy-intensive.
To overcome this, the team developed machine learning algorithms trained on sharp braille images with artificially added blur. After learning to ‘deblur’ these images, they applied a computer vision model for character detection and classification.
The resulting robotic braille reader, capable of quickly scanning rows of braille characters, demonstrated a reading speed and accuracy comparable to human readers. Hardman found the accuracy surprising given the use of artificially blurred images for training.
Potdar suggests that braille reading speed is an excellent metric for evaluating tactile sensing systems, potentially impacting areas beyond braille reading, such as surface texture detection and slip prevention in robotics.
Looking forward, the researchers aim to scale this technology to the size of a humanoid hand or skin, with partial funding support from the Samsung Global Research Outreach Program.
покупка аккаунтов маркетплейс аккаунтов соцсетей
магазин аккаунтов социальных сетей продажа аккаунтов
Account Buying Platform Account Acquisition
Account Trading Platform Social media account marketplace
Accounts for Sale Account Buying Service
Online Account Store Account Purchase
account marketplace sell pre-made account
gaming account marketplace accounts marketplace
account selling service ready-made accounts for sale
database of accounts for sale accounts market
account market website for selling accounts
account exchange buy pre-made account
gaming account marketplace account store
account market account trading
buy accounts account trading service
account store account trading platform
purchase ready-made accounts sell accounts
online account store buy accounts
buy pre-made account ready-made accounts for sale
sell accounts accounts marketplace
account buying service https://social-accounts-marketplaces.live
online account store accounts market