Robots Get a Grip

By and large, robotic end effectors take the form of some type of gripper. These grippers seek, to varying degrees, to reproduce the function of human fingers. This is a useful arrangement for many applications, but it also misses the fact that human hands are more than just fingers. When we tightly grasp an object the palm is a key factor in strengthening our grip. The palm can also come into play in situations where we, for example, rotate an object with a single hand.

Ignoring the complexities of hand anatomy can make end effector development and control much simpler. However, when human-like capabilities are needed, these simplified implements may not be able to get the job done. For use cases such as these, many aspects of the palm need to be reproduced accurately. This includes the ability to bend to reposition the fingers, envelop an object that comes into contact with it, and sense these contact events with a high degree of precision.

Existing artificial systems do not fully meet present needs for human-like grasping, sensing, and manipulation, so a pair of researchers at MIT put forth a new solution. Their approach involved the development of a soft palm, called GelPalm, with embedded sensors that can conform to the shape of objects that it comes into contact with. This palm was paired with RObotic Modular Endoskeleton Optical (ROMEO) fingers, which are outfitted with a similar sensing system, to complete the robotic hand design.

Both GelPalm and ROMEO have a sensing system that was inspired by GelSight sensors. They consist of a soft, flexible silicone “skin” that covers a rigid, jointed skeletal structure. Inside of the skin, a set of colored LEDs illuminate the inner surfaces. A camera then captures the light that is reflected back from those surfaces. As the fingers or palm come into contact with external objects, the skin is deformed, which causes the reflected light to scatter in predictable ways. A processing algorithm can then determine exactly where, and how firmly, contact with objects is being made.

The rigid skeleton of the ROMEO fingers was 3D-printed to make the fabrication process simple and inexpensive. The finger joints have artificial tendons attached to them, which cause the fingers to actuate when they are pulled on by a motor hidden in the base of the hand. To minimize the complexity and size of the device, the palm requires no actuators and instead adopts a passive compliance strategy. It was designed such that if force is applied along its length, it will deform to grip the object.

Some tests were conducted in which objects with wet paint were grasped by the team’s robot hand, and also by other versions that are less compliant. It was found that the team’s design gave the largest surface contact areas, which indicates that it can maintain a better grasp on a variety of items.

In the future, the team plans to streamline their design to minimize its complexity, and also to experiment with other types of fingers, which could, for example, allow the hand to grasp larger objects. This work could lead to the development of many new applications in prosthetics, manufacturing, and beyond.