Article

ORD-WM: Object-Centric Rigid Dynamics 3D World Model for Robotic Manipulation

ORD-WM proposes a 3D world model that learns object-centric rigid-body dynamics for robotic manipulation tasks. It uses a vector-neuron variational autoencoder supervised by signed distance fields to encode each object point cloud into a geometry-aware latent representation composed of 3D vector features, and uses an action-conditioned dynamics model to predict per-object SE(3) motion increments. It then analytically applies the predicted rigid-body transformations to the current latent state rather than generating an unconstrained next latent state, thereby separating persistent geometry from pose changes.

Part of the event: ORD-WM proposes object-centric rigid-body dynamics 3D world model

This page provides a summary and source link only.