One area of significant importance in human-robot interaction (HRI) is the ability of robots to grasp objects. The robotic manipulation of items is a challenging task that requires various sensors, non-traditional grippers, and complex algorithms. In addition, the shape and size of the flat object can highly influence this process. 

We developed an HRI framework using MR for object manipulation without requiring conventional sensors and specialized grippers. To the best of our knowledge, this is the first time addressing the problem of object manipulation without using external sensors and dexterous grippers but an MR headset (MRH). In addition, we introduce a deep reinforcement learning technique (DRL), such as proximal policy optimization (PPO).