alt text 

Stanford University Networking Seminar


Nan Tian (UC Berkeley)
A Cloud-and-Edge Hybrid Approach for Dynamic Human Robot Interactions

12:15pm, Thursday, May 10 2018
Gates 104

About the Talk

With Cloud computing and fast networks, robots are no longer limited by onboard computation powers. Large scale deployments of robots can benefit from a centrally maintained robotic-automation-as-a-service (RaaS) framework. We have demonstrated this concept previously by deploying Dex-net in the cloud as a grasping recommendation system to guide a dual arm robot to play non-standard “wizard” chess.

However, in service robots, highly dynamic perception as well as interaction between the robot and the humans are necessary. In this talk, we propose a hybrid approach to combine the power of the cloud and the speed of the “edge” controllers for a more agile system. We demonstrate how to use it for a range of human robot interaction tasks, from gesture followings to dynamic object pickups, all programmed in a cloud framework called HARI.

About the Speaker

Nan Tian is a PhD student in the UC Berkeley CS department. His current research includes cloud robotics and visual feed back based hand manipulation for service robots. Previously trained as biomedical engineer in neural science and medical imaging, he has discovered his new passion in high performance computing, computer vision, and robotics after joining UC Berkeley.