A Compositional Object-Based Approach to Learning Physical Dynamics
This work addresses the challenge of building intuitive physics simulators for AI systems, offering a novel approach that could enhance generalization in robotics or virtual environments, though it is incremental as it builds on existing neural and symbolic methods.
The paper tackles the problem of learning physical dynamics by introducing the Neural Physics Engine (NPE), a framework that uses compositional object-based representations to predict movement and generalize across variable object counts and scene configurations, showing improved performance in predicting movement and inferring latent properties like mass compared to less structured architectures.
We present the Neural Physics Engine (NPE), a framework for learning simulators of intuitive physics that naturally generalize across variable object count and different scene configurations. We propose a factorization of a physical scene into composable object-based representations and a neural network architecture whose compositional structure factorizes object dynamics into pairwise interactions. Like a symbolic physics engine, the NPE is endowed with generic notions of objects and their interactions; realized as a neural network, it can be trained via stochastic gradient descent to adapt to specific object properties and dynamics of different worlds. We evaluate the efficacy of our approach on simple rigid body dynamics in two-dimensional worlds. By comparing to less structured architectures, we show that the NPE's compositional representation of the structure in physical interactions improves its ability to predict movement, generalize across variable object count and different scene configurations, and infer latent properties of objects such as mass.