Sushi robot gets a fancy sim trainer

A robot with a sharp knife sounds scary enough, but now there is a simulator for efficient robotic cutting that predicts the forces acting on a knife as it slices through foods such as fruits and vegetables. Sushi could be next on the chopping block.

Researchers from Nvidia and University of South California presented the concept for DiSECt, a differentiable simulator for robotic cutting, at the 2021 Robotics: Science and Systems conference.  Their work is summarized on a new Nvidia Developer blog.

There is also a 4 minute video describing their work:

The blog describes the complex cutting process captured in the simulation, noting “no two cucumbers or tomatoes are the same.”  The researchers got around this problem with a new way of cutting which envisions the potato or other object in a 3D mesh, then applying an algorithm that duplicates the mesh elements with virtual vertices on the edges where cuts are to be made.

Then, DiSECt inserts virtual springs on either side of the cutting surface so it is possible to simulate damage mechanics and crack propagation continuously. “We can efficiently determine an energy-minimizing yet fast-cutting motion,” according to the blog by Eric Heiden.

Ongoing work by the research team uses differentiable simulation to perform real-world robotic cutting in which the simulator is updated online from measurements of force while a real-life robot is actually cutting through food such as potatoes and apples. The researchers hope to energy efficient and time efficient ways to apply cutting actions to actual robots.

RELATED: From cuddly seals to AVs, robotics are poised for growth