MIT’s new robot gently peels squash with one hand while holding it by other

Researchers at MIT have developed a new robot that can easily perform some of type of kitchen tasks such as peeling multiple vegetables.

In a recent video, the robot is seen peeling a squash with one hand while holding it by the other.

Using its finger, the robotic system can easily rotate fruits and vegetables and use its other hand to peel them.

Robot addresses dexterous manipulation problems

Recent studies have made significant progress in addressing dexterous manipulation problems, particularly in in-hand object reorientation.

However, there are few existing works that explore the potential utilization of developed dexterous manipulation controllers for downstream tasks.

In this new study, MIT researchers focused on constrained dexterous manipulation for food peeling. Food peeling presents various constraints on the reorientation controller, such as the requirement for the hand to securely hold the object after reorientation for peeling.

“We propose a simple system for learning a reorientation controller that facilitates the subsequent peeling task,” said researchers.

Humans peel potatoes by grasping the potato in one hand and using the second hand to actuate a peeler to remove the potato’s skin. After a part of the potato is peeled, it is rotated while being held in the hand (i.e., in-hand manipulation) and peeled again. The sequence of rotating and peeling continues until all of the potato’s skin is removed.

Robotic system can re-orient different vegetables

“In this work, we present a robotic system that can re-orient different vegetables using an Allegro hand in a way that their skin can be peeled using another manipulator,” said researchers .

The dexterous manipulation system that utilizes an Allegro hand mounted on a Franka robot arm to reorient food items for downstream peeling. The other Franka robot arm uses its gripper to grasp a peeler for peeling.
The reorientation controller for the Allegro hand is learned through reinforcement learning, while the peeling is performed via teleoperation.

Researchers demonstrated the process of reorienting and peeling a melon, a sweet potato, and a squash.

The reorientation controller presented in this study is a blind controller that relies solely on proprioceptive sensory information. While it has demonstrated the ability to successfully reorient heavy objects and securely hold them in place, its performance could potentially be enhanced by incorporating visual and tactile feedback.

“These additional steps of doing rotation are something which is very straightforward to humans, we don’t even think about it,” Pulkit Agrawal, assistant professor in the department of Electrical Engineering and Computer Science (EECS) at MIT, told New Scientist . “But for a robot, this becomes challenging.”

The current system has a few failure modes. Firstly, the object might slip out of the hand since the controller does not utilize any vision information. Secondly, the controller might fail if the vegetables are small, as the fingers cannot effectively make contact with the object.