Vision-based 3D object localization for manufacturing applications

The industrial scale manufacturing automation largely works because of robotic systems which are preprogrammed to reliably carry out the same sequence of steps over and over again. As a result, when scenarios change, or become more complex, robotic systems need to be re-programmed. Moreover, in high value added manufacturing applications, it is impractical to program every single motion of a robot manually. The project aims to develop fully autonomous robotic manufacturing and inspection systems that do not require extensive programming by end-users. In particular, a key step in automation of manufacturing is the ability to accurately detect and localize parts so that a robot gripper can accordingly adjust. This project will capitalize on recent developments in modern AI to develop automated methods for accurate localization of objects through images captured by a camera.

Mahboobeh Norouzi
Faculty Supervisor: 
Ahmed Ashraf
Partner University: