A team of robotics engineers has created a model that predicts the force needed to push against various fixtures in the environment by a robotic gripper in order to adjust its grasp on an object.
Most robots on a factory floor are equipped with large claws or pincers. They are fairly ham-handed and designed only to perform simple maneuvers, such as grabbing an item and move it somewhere else in an assembly line. However, more complex movements are still out of reach for many industrial robots, including adjusting the grasp on an object.
Engineers at Massachusetts Institute of Technology (MIT) have found a way to provide simple robotic grippers with more dexterity by making use of their environment. The team, led by an assistant professor of mechanical engineering, Alberto Rodriguez and Nikhil Chavan-Dafle, a graduate student, has developed a model that predicts the force with needed by a robotic gripper to push against various fixtures in the environment.
For example, if a robotic gripper tries to pick up a pencil at its midpoint, it could use the environment to adjust its grasp by grabbing hold of the eraser end Instead of releasing the pencil and trying again. A robot using Rodriguez's model can loosen its grip slightly and push the pencil against a nearby wall, then slide the gripper closer to the pencil's midpoint.
Making robots to partner with the environment in order to improve their dexterity is an approach called by Rodriguez "extrinsic dexterity". This is opposed to the intrinsic dexterity of the human hand. A person would simply use one hand to adjust grip on a pencil in a similar fashion, by simply spider-crawl fingers towards the center of the pencil. However, it is extremely difficult to program such intrinsic dexterity in a robotic hand and the complex software would significantly raise a robot's cost.
Existing robots in disaster response, medicine, manufacturing and other gripper based applications may interact with their environment in a simple and cost-effective way with Rodriguez's new approach. This can allow them to perform more complex maneuvers.
The study will be presented by Rodriguez and Chavan-Dafle at the International Conference on Intelligent Robotics and Systems in September. The National Science Foundation was supporting, in part, their research.
