Scene Interaction

Scene interaction research focuses on realistically simulating and understanding how humans and robots interact with their environments. Current efforts leverage machine learning, particularly reinforcement learning and transformer-based generative models, to synthesize natural human-scene interactions from diverse data sources, often incorporating physics-based simulations and semantic scene understanding. This work is driven by applications in animation, robotics, and assistive technologies for the visually impaired, aiming to create more realistic virtual characters and enable robots to autonomously explore and map physical properties of their surroundings.