MVISTA-4D: View-Consistent 4D World Model with Test-Time Action Inference for Robotic Manipulation

TL;DR

This paper introduces MVISTA-4D, a novel 4D world model for robotic manipulation that generates consistent multi-view RGBD sequences from a single view and infers actions through test-time optimization, improving scene understanding and manipulation accuracy.

Contribution

It presents a new embodied 4D world model with view-consistent RGBD generation and a test-time action inference method, advancing scene prediction and robotic control.

Findings

Strong performance on 4D scene generation tasks

Effective action inference via test-time optimization

Improved manipulation accuracy across datasets

Abstract

World-model-based imagine-then-act becomes a promising paradigm for robotic manipulation, yet existing approaches typically support either purely image-based forecasting or reasoning over partial 3D geometry, limiting their ability to predict complete 4D scene dynamics. This work proposes a novel embodied 4D world model that enables geometrically consistent, arbitrary-view RGBD generation: given only a single-view RGBD observation as input, the model imagines the remaining viewpoints, which can then be back-projected and fused to assemble a more complete 3D structure across time. To efficiently learn the multi-view, cross-modality generation, we explicitly design cross-view and cross-modality feature fusion that jointly encourage consistency between RGB and depth and enforce geometric alignment across views. Beyond prediction, converting generated futures into actions is often handled by inverse dynamics, which is ill-posed because multiple actions can explain the same transition. We address this with a test-time action optimization strategy that backpropagates through the generative model to infer a trajectory-level latent best matching the predicted future, and a residual inverse dynamics model that turns this trajectory prior into accurate executable actions. Experiments on three datasets demonstrate strong performance on both 4D scene generation and downstream manipulation, and ablations provide practical insights into the key design choices.