Communications of the ACM
Deep Learning Helps Robots Grasp, Move Objects with Ease
Artificial intelligence software developed at the University of California, Berkeley integrates neural networks with motion-planning software to enable robots to assist humans in warehouse settings.
Credit: Ken Goldberg Lab
University of California, Berkeley (UC Berkeley) engineers have developed artificial intelligence software that integrates neural networks with motion-planning software to enable robots to assist humans in warehouse settings.
The software gives robots the speed and skill to grip and smoothly move objects.
The UC Berkeley researchers upgraded a Grasp-Optimized Motion Planner that can compute how a robot should grip an object and move to transfer it between locations. They added a deep learning neural network, accelerating the planner's computing time and eliminating jerky movements.
The combined neural network and motion planner shortened average computation time from 29 seconds to 80 milliseconds.
From Berkeley News
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found