Get a Grip!

from hackster.io

The ability to grip objects is a basic skill that robots need to learn in order to interact with their environments in varied ways. The fact that this skill comes so easily to humans can make it seem deceptively simple to implement mechanically. But when considering the variety of objects — varying in size, weight, texture, fragility, and so on — that a robot will encounter in the real world, the task gets complicated very quickly. As such, developing a gripper that can perform well in a large number of use cases has been an elusive goal to date.

The latest to throw their hat in the ring is a group of engineers from the National University of Singapore, with their hybrid robotic grippers that were developed with commercial applications in mind. The grippers make use of soft, flexible 3D-printed fingers and a reconfigurable gripper base that allow the device to adapt to many situations, including food assembly, vertical farming, and fast-moving consumer goods packaging.

Many industries still rely on humans to handle delicate items because of the difficulty of designing a robotic system to tackle the task. This research team instead applied computer vision and deep learning to the problem of recognizing object types and orientations. With that data, the gripper has enough information to determine how best to pick and place the detected object, regardless of shape, size, fragility, and stiffness.

The grippers themselves consist of three or four soft fingers, which are reconfigurable. Each finger is individually actuated by air flow, and has a novel locking mechanism to adjust stiffness. In total, three variants of this gripper have been developed, which allow it to be used under different circumstances.

One variant, called GourmetGrip, was designed for delicate tasks like handling soft foods. It has been shown to be capable of packing foods as soft as tofu into take-out boxes, and to do so at speeds comparable to what humans can achieve. When compared with other commercially available robotic grippers, GourmetGrip has been shown to be more accurate, and to operate approximately 23 percent faster. The gripper has already been proven to be effective with over fifty foods, including foods as soft as cake and pudding.

UnisoGrip, a second variant, is applicable to a wider range of tasks. It was designed with the goal of placing packaged goods into boxes for shipping and transportation. UnisoGrip can expand its grip very wide to accommodate large objects, and also has a vacuum suction cup for awkwardly positioned objects. In handling over thirty types of consumer goods, including bottled drinks, coffee powder packs, and refillable detergent packs, UnisoGrip was found to have a 20 percent higher gripping efficiency than other grippers currently on the market.

The third gripper, based on the GourmetGrip and UnisoGrip platforms, is highly customizable such that it can adapt to many use cases and space requirements. This gripper has already been deployed in a Singapore-based factory where it packs rice vermicelli packets into carton boxes. This installation has been able to reduce the workload of existing staff members, while simultaneously boosting their productivity. That combination of benefits may just propel this gripper system into other industrial applications in the future.