Object Contact

Object contact research focuses on understanding and modeling how humans and robots interact with objects physically, aiming to improve robotic manipulation, human-computer interaction, and 3D scene generation. Current research employs various approaches, including deep reinforcement learning for robotic control, diffusion models for 3D scene generation incorporating spatial constraints and collision avoidance, and transformer networks for joint 3D human-object reconstruction leveraging contact information. These advancements are crucial for creating more realistic virtual environments, improving robotic dexterity and safety, and enabling more natural human-robot collaboration.