Customer name：UC Berkeley
Application type：Robot control and analysis
Project time：May 2020
Application product：Qinglong 2
Application type：Deep learning, neural network
Project Overview：Researchers at the University of California, Berkeley have developed new artificial intelligence software that enables robots to quickly and skillfully grasp and move objects smoothly, allowing them to help people in warehouse environments in the near future. The calculation time of the motion planner is greatly shortened by integrating the deep learning neural network. Neural networks allow robots to learn from examples. Later, robots can usually be generalized to similar objects and movements.
Project results： The results were published in the "Science Robot" magazine. By combining the neural network with the exercise planner, the team reduced the average calculation time from 29 seconds to 80 milliseconds, or less than a tenth of a second.