An estimated 68,000,000 tons of garbage is sifted through recycling companies every year. The process involves a fast-moving conveyor belt that carries trash that needs to be sorted for recycling.

Trash is sorted into different categories such as paper, plastic, and glass. During the process ofsorting, the workers are tasked to act as quickly as they can to manually separate non-recyclable items, contaminated trash, or those that belong to a different category.

The task is tedious, repetitive, dirty, and at times, dangerous as working trash can sometimes mean that glass, wires, or other sharp objects are mixed in with the garbage.

With this in mind, a group of researchers from Massachusetts Institute of Technology's (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed a robotic hand that will aid in the recycling process.

The team called the project RoCycle which is composed of a soft Teflon hand with tactile sensors on its fingertips, which can detect various physical properties of an object such as size, stiffness, and material among other things. The team designed RoCycle to be compatible with any robotic arm. Currently, it has 85% accuracy in detecting stationary materials, and 63% when items are on a conveyor belt, which is the most common error when it comes to machines designed for recycling.

MIT professor Daniela Rus explains that the robot has sensorized skin that provides haptic feedback, which in turn, allows the robot to determine to which category an item belongs. Rus explains that computer vision is not enough to perform the task and tactile input is one important source of info information for refined results.
The team has collaborated with Yale University resulting in positive results such as getting the robot to differentiate a paper cup from a plastic cup with similar appearances. This feature is an advancement as compared to other models with vision systems that would have trouble with the said task.

The project is a step closer to the larger goal of Prof. Rus which is to reduce the back-end cost of recycling. This could incentivize more cities, or even countries, to create and implement programs that would help reduce solid waste and promote recycling.

Today's current technology in recycling is equipped with different features that have its respective tasks. Such features include optical sorters that use different wavelengths of light to distinguish the difference between types of plastics. There are also aluminum sorters that use eddy currents which removes items with non-magnetic metals. Some machines also feature magnetic sorters that separate iron and steel products.

The lead author of the paper, a Ph.D. student named Lillian Chin, explains that if RoCycle would be deployed on a wider scale the single-stream recycling will have lower contamination rates than that of multi-stream recycling.

Developing a machine that can distinguish between paper, plastic, and metal is a highly difficult endeavor, making this project a very impressive feat. Programming human-like tactile recognition is difficult to program into robots.

To further improve the accuracy of RoCycle, the researchers are planning to integrate actual video data from the robot's cameras. This will potentially avoid problems in differentiating trash items.