Tactile Sensor Gives Robots Novel Dexterity
Using modified high-end sensors, scientists at Northeastern University and MIT have built a robot with unprecedented mobility in its hands.
The sensor, which was developed using an existing technology called GelSight, enabled the robot to reach for a dangling USB cable and plug it into the a USB port.
Like Us on Facebook
Prof. Robert Platt, of the Computer Science department at Northeastern University said robots are generally capable of moving objects when these object are stationary.
A robot's ability to "take its bearings as it goes," is unheard of, Platt pointed out.
Platt said that for a long time, researches have been trying to develop the kind of technology the sensor provides, but were faced with failure because of the sensor's inaccuracy and insufficiency of information.
The GelSight technology was first described in 2009 and developed by John and Dorothy Professor of Vision Science, Edward Adelson of the Massachusetts Institute of Technology.
The new sensor, however, is less sensitive than GelSight, and was developed with a new algorithm that made it faster. The sensor can also provide a real-time feedback to the robot.
In addition, the new sensor is also smaller than GelSight, which makes it possible for researchers to implant it in a robot's hands.
The GelSight sensor uses a computer-vision algorithm and optics.
Both the original and the newly developed sensor have a synthetic and transparent rubber covering painted with a metallic coating on one side.
The rubber is able to adapt with any surface it's pressed against, while the metallic side enables the sensor to reflect light.
Adelson said he became interested in tactile sensors in robotics because he knew it would fascinate his children. Additionally, he said he was interested in how robots use their visual system to move their fingers.
"Since I'm a visual guy, if it's an image, I know what to do with it," Adelson said.
Researchers presented the sensor during the International Conference on Intelligent Robots and Systems last week.
The MIT team that developed the sensor included Adelson, PhD student Rui Li, master's student Wenzhen Yuan and senior research scientist Mandayam Srinivasan from the Department of Mechanical Engineering.
Meanwhile, the Northeastern University team that conducted the experiments on the sensor and built its controller included Platt, Nathan Roscup and Andreas ten Pas.