In this work, I explore the task of robot grasping of dynamic objects in dynamic scenarios. This involves reasoning about how to grasp the object, how to approach it while avoiding obstacles in the environment, and what actions to take in case of occlusion. I explored deep learning-based methods for grasp generation, such as Contact GraspNet and AnyGrasp. For path planning, we utilize the Vector Accelerated Motion Planning library, and for path execution, we use the Universal Robotics Real-Time Data Exchange library.
Demo Video:
Demo.Video.mp4
High Quality Video link: Grasp-It Demo
Project Poster:
