Researchers from the University of California, Berkeley (UC Berkeley) have improved the time it takes for sorting robots to grasp and move objects using artificial intelligence (AI).

Working with previously developed software called Grasp-Optimized Motion Planner — which determines how a robot will pick up and transfer an object — the UC Berkeley team incorporated a deep learning neural network into the software.

Deep learning — which is a function of AI that mimics how the human brain processes data for use in object detection, speech recognition, language translation and decision making — expedited the “jerky” moves of the Grasp-Optimized Motion Planner from an almost 30 second process to a smoother process that takes just one-tenth of a second.

The UC Berkeley team believes that the technology will expedite processes in manufacturing and warehouse settings, enabling robots to quickly grasp and transfer objects.

The research is detailed in the journal Science Robotics and in the accompanying video that appears courtesy of UC Berkeley.