Peter Battaglia

Google (United States), Google DeepMind (United Kingdom)

Papers

Total Citations

330

H-Index

About

Peter Battaglia is a leading researcher at the intersection of artificial intelligence, physics, and computer graphics, whose work has fundamentally reshaped how machines perceive and simulate the physical world. His key contributions lie in developing neural network architectures that can learn intuitive physics directly from visual data. Battaglia pioneered the concept of "Visual Interaction Networks," which allow AI systems to infer the underlying dynamics of physical systems—such as collisions and gravitational forces—from raw video, achieving over 250 combined citations for his seminal 2017 papers. This work bridged the gap between human-like physical intuition and machine learning. More recently, Battaglia has advanced generative modeling of 3D geometry with "PolyGen" (65 citations), an autoregressive model that directly outputs polygon meshes, a critical tool for robotics and graphics. His research on "face interaction graph networks" (2022) further extends graph neural networks to simulate complex rigid-body dynamics, tackling notoriously difficult problems in engineering and animation. Through these innovations, Battaglia has established himself as a pivotal figure in creating AI systems that can reason about, predict, and generate the physical structure of our world.