Unsupervised Learning of Latent Physical Properties Using Perception-Prediction Networks
This work addresses the challenge of discovering relevant object properties in dynamical systems without supervision, which could benefit fields like robotics and physics simulation, though it appears incremental as it builds on existing graph-based neural architectures.
The authors tackled the problem of unsupervised learning of latent physical properties from object interactions, achieving accurate simulation of dynamics for unseen objects and translating learned representations into human-interpretable properties like mass and coefficient of restitution.
We propose a framework for the completely unsupervised learning of latent object properties from their interactions: the perception-prediction network (PPN). Consisting of a perception module that extracts representations of latent object properties and a prediction module that uses those extracted properties to simulate system dynamics, the PPN can be trained in an end-to-end fashion purely from samples of object dynamics. The representations of latent object properties learned by PPNs not only are sufficient to accurately simulate the dynamics of systems comprised of previously unseen objects, but also can be translated directly into human-interpretable properties (e.g., mass, coefficient of restitution) in an entirely unsupervised manner. Crucially, PPNs also generalize to novel scenarios: their gradient-based training can be applied to many dynamical systems and their graph-based structure functions over systems comprised of different numbers of objects. Our results demonstrate the efficacy of graph-based neural architectures in object-centric inference and prediction tasks, and our model has the potential to discover relevant object properties in systems that are not yet well understood.