U4D: Uncertainty-Aware 4D World Modeling from LiDAR Sequences

TL;DR

U4D introduces an uncertainty-aware 4D LiDAR modeling framework that improves realism and temporal stability by focusing on uncertain regions and ensuring coherence in dynamic 3D scene reconstruction.

Contribution

The paper proposes a novel uncertainty-aware approach with a two-stage generation process and a spatio-temporal fusion block for more accurate and stable 4D LiDAR scene modeling.

Findings

Enhanced geometric fidelity in complex regions

Improved temporal consistency of LiDAR sequences

Outperforms existing methods in realism and stability

Abstract

Modeling dynamic 3D environments from LiDAR sequences is central to building reliable 4D worlds for autonomous driving and embodied AI. Existing generative frameworks, however, often treat all spatial regions uniformly, overlooking the varying uncertainty across real-world scenes. This uniform generation leads to artifacts in complex or ambiguous regions, limiting realism and temporal stability. In this work, we present U4D, an uncertainty-aware framework for 4D LiDAR world modeling. Our approach first estimates spatial uncertainty maps from a pretrained segmentation model to localize semantically challenging regions. It then performs generation in a "hard-to-easy" manner through two sequential stages: (1) uncertainty-region modeling, which reconstructs high-entropy regions with fine geometric fidelity, and (2) uncertainty-conditioned completion, which synthesizes the remaining areas under learned structural priors. To further ensure temporal coherence, U4D incorporates a mixture of spatio-temporal (MoST) block that adaptively fuses spatial and temporal representations during diffusion. Extensive experiments show that U4D produces geometrically faithful and temporally consistent LiDAR sequences, advancing the reliability of 4D world modeling for autonomous perception and simulation.