Assembly line robots are becoming more sophisticated but they are still in the Stone Age when it comes to the dexterity of their hands. Mechanical pincers and claws are no match for the deftness of the human hand. Now engineers at MIT have found a way to make industrial robots nimbler without the need for huge investments in complex robotic limbs.
The key is to use the external environment to imbue mechanical grippers with more dexterity.
Consider, for example, a mechanical gripper that picks up a pencil by grasping the eraser at the end rather than the midpoint where its hold is much firmer. Instead of releasing the pencil and trying again, the machine loosens its grip, and slides the gripper closer to the midpoint by pushing the pencil against a surface such as a wall.
This seemingly simple method for increasing dexterity could enable industrial robots to carry out a wider range of tasks and become more versatile.
The MIT team of engineers developing the method is led by Alberto Rodriguez, Assistant Professor of Mechanical Engineering, and graduate student Nikhil Chavan-Dafle. They have created a model that predicts the force with which a robotic gripper needs to push against various fixtures in the environment in order to adjust its grip on an object. They call it “extrinsic dexterity.”
There are numerous ways in which the method can be applied, and the team is now exploring a number of options. These include using gravity to toss and catch an object, and using large, flat surfaces such as tabletops to help a robot roll an object between its fingers. Corners and edges are other external features that could be employed in this way.
The idea might seem simple – especially in terms of the human hand’s intricate engineering – but it is actually difficult to execute. In order to predict how an object might move as a gripper pushes it against a fixture, the model has to take into account various factors. These include the frictional forces between the gripper and the object and between the object and the environment, as well as the object’s mass, inertia, and shape. The model can calculate the force a gripper must exert to maneuver an object to a desired orientation; the amount of push required to tilt the item to an angle of 45 degrees, for example.
The researchers validated the model by manipulating objects with a simple, two-fingered gripper, and comparing the actual forces involved against those predicted by the model. Now the team is looking at how to plan motions to generate certain trajectories. One question they will address is how fixtures in the environment can be engineered to make a robot’s motions more reliable and quicker.
Ultimately, Rodriguez sees extrinsic dexterity as an inexpensive way to make simple robots more nimble for a variety of uses, including applications in manufacturing.
This post is based on an article published in MIT News. Access the full article here.