Automating warehouse tasks can be challenging. Many actions that come naturally to people are actually
“Warehouses are still serviced mainlyhumans, because robots still have a very difficult time reliably capturing many different objects, explains Ken Goldberg, senior author of the study. - On an automotive assembly line, the same movement is repeated over and over, so it can be automated. But in the warehouse, all orders are different. "
Video demonstrating the work of a warehouse robotic arm with movement before and after the application of a deep neural network. Credit: Ichnowski et al.
In earlier work, Goldberg and the scientificUC Berkeley researcher Jeffrey Ichnovsky created a motion planner optimized for grasping. However, the engineers were unable to make the movements smooth. Although the software parameters were adjusted to create exactly this, the calculations took on average about half an hour.
In a new study by Goldberg and Ichnowski inCollaboration with UC Berkeley graduate student Yahav Avigal and student Vishal Satish significantly accelerated the computation time of the motion planner by integrating a neural network with deep learning.
Neural networks allow the robot to learn fromexamples. Later, the robot can often generalize similar objects and movements. However, these approximations are not always accurate enough. Goldberg and Ichnovsky found that the approximation generated by the neural network could then be optimized using a custom scheduler. By combining the neural network with the motion planner, the team reduced the average computation time from 29 seconds to 80 milliseconds.
The authors of the study are confident that thanks to thisand other advances, robots will soon be able to help warehouse workers. “This is a great new opportunity for robots to support people,” Goldberg concludes.
Read also
Neurons in the human brain and the network of galaxies are similar
Due to plate movement, the Pacific Ocean floor is now deep under China
Scientists hacked a robot vacuum cleaner to remotely record user conversations