Robots that can peel and slice curvy veggies

Thursday, 23 April, 2026

Humans can routinely interact with curved objects as part of daily life, for example, when slicing a banana, peeling a cucumber or washing dishes. However, handling curvy objects can be challenging for robots. Now, researchers have developed an approach that can help robots transfer skills across objects of diverse shapes and sizes, enabling them to peel and slice varied fruits and vegetables.

In order for robots to manipulate an object, they need to understand its position, shape and proportions. While planar or regular shapes share a global frame that can be represented by common reference points, it is much more difficult to create a generalised frame for curved objects. This, in turn, makes it harder for robots to transfer actions across different curved objects.

To address these challenges, Cem Bilaloglu and colleagues designed a way to generate reference shapes for irregular objects, thus allowing robots to generalise tasks regardless of geometry.

This method generates a point cloud of coordinates for the object and then defines key reference points along its surface.

These points can form a smooth representation of the object that includes task-relevant information, which can then be combined with control or trajectory-planning frameworks.

Using this method, the researchers demonstrated that robots could perform continuous, contact-rich actions on several everyday objects.

Tasks included peeling a banana or sweet potato, slicing vegetables, and probing the surface of different 3D-printed objects. Bilaloglu et al. also showed that their approach could operate even with partial or noisy sensor data, or in cluttered workspaces.

The study was published in Science Robotics.