Learning Generalizable Physical Dynamics of 3D Rigid Objects
This addresses the problem of generalizable physical prediction for applications like robotics and autonomous vehicles, representing an incremental advance by focusing on shape generalization.
The paper tackles predicting the dynamics of 3D rigid objects, such as final resting position and total rotation under impulsive forces, achieving accurate predictions for unseen object shapes using a neural network trained on physics engine data.
Humans have a remarkable ability to predict the effect of physical interactions on the dynamics of objects. Endowing machines with this ability would allow important applications in areas like robotics and autonomous vehicles. In this work, we focus on predicting the dynamics of 3D rigid objects, in particular an object's final resting position and total rotation when subjected to an impulsive force. Different from previous work, our approach is capable of generalizing to unseen object shapes - an important requirement for real-world applications. To achieve this, we represent object shape as a 3D point cloud that is used as input to a neural network, making our approach agnostic to appearance variation. The design of our network is informed by an understanding of physical laws. We train our model with data from a physics engine that simulates the dynamics of a large number of shapes. Experiments show that we can accurately predict the resting position and total rotation for unseen object geometries.