Advances in AI and computer vision have allowed robots to gain a better understanding of their environment and accomplish complex manipulation tasks. However, traditional robotic systems struggle with occluded environments, where visual sensors provide limited information. In my project, I aim to create a first-of-its-kind robotic grasping system capable of grasping objects in fully occluded environments (e.g., under packaging materials). I will build on recent work which uses millimeter-wave (mmWave) signals to create 3D reconstructions of hidden objects and investigate new machine learning architectures to predict optimal grasp points for this new modality. This work can open up new possibilities in robotic perception and manipulation, with potential uses in warehouses and smarthomes.

My goals for participating in this SuperUROP are to gain high-level research experience in preparation for an MEng, and to deepen my knowledge in machine learning (ML), sensing, and robotic manipulation. With my project, I am able to regularly apply skills that I had learned from my courses, such as ROS and ML. Ultimately, I am motivated by the potential for my research to advance robotic applications in both domestic and industrial settings.