← World Models

LiDAR Generation

Leveraging point cloud sequences from LiDAR sensors to generate geometry-grounded scenes for safety-critical domains.

⌘K

LiDAR generation world models synthesize and predict point cloud sequences from LiDAR sensors, enabling geometry-grounded scene generation, future forecasting, and autoregressive simulation for autonomous driving and safety-critical applications.

LiDAR Data Engines — Point Cloud Generation & Completion 30+ papers

Model/MethodFull TitleVenueYear
DUStyLearning to Drop Points for LiDAR Scan Synthesis IROS2021
LiDARGenLearning to Generate Realistic LiDAR Point Clouds ECCV2022
UltraLiDARLearning Compact Representations for LiDAR Completion and Generation CVPR2023
Copilot4DLearning Unsupervised World Models for AD via Discrete Diffusion ICLR2024
R2DMLiDAR Data Synthesis with Denoising Diffusion Probabilistic Models ICRA2024
ViDARVisual Point Cloud Forecasting enables Scalable Autonomous Driving CVPR2024
LiDiffScaling Diffusion Models to Real-World 3D LiDAR Scene Completion CVPR2024
LiDMTowards Realistic Scene Generation with LiDAR Diffusion Models CVPR2024
RangeLDMFast Realistic LiDAR Point Cloud Generation ECCV2024
Text2LiDARText-Guided LiDAR Point Cloud Generation ECCV2024
LiDARGRITTaming Transformers for Realistic LiDAR Point Cloud Generation arXiv2024
BEVWorldMultimodal World Simulator for AD via Scene-Level BEV Latents arXiv2024
LOGenToward LiDAR Object Generation by Point Diffusion arXiv2024
OLiDMObject-Aware LiDAR Diffusion Models for Autonomous Driving AAAI2025
X-DriveCross-Modality Consistent Multi-Sensor Data Synthesis ICLR2025
LidarDMGenerative LiDAR Simulation in a Generated World ICRA2025
LiDAR-EDITLiDAR Data Generation by Editing Object Layouts ICRA2025
R2FlowFast LiDAR Data Generation with Rectified Flows ICRA2025
WeatherGenA Unified Diverse Weather Generator for LiDAR Point Clouds CVPR2025
SuperPCSingle Diffusion Model for Point Cloud Completion, Upsampling, Denoising CVPR2025
HERMESA Unified Self-Driving World Model for 3D Scene Understanding and Generation ICCV2025
SPIRALSemantic-Aware Progressive LiDAR Scene Generation and Understanding NeurIPS2025
DiffSSCSemantic LiDAR Scan Completion using Denoising Diffusion IROS2025
RadarGenAutomotive Radar Point Cloud Generation from Cameras arXiv2025
OpenDWMOpen Driving World Models arXiv2025
DriveXOmni Scene Modeling for Learning Generalizable World Knowledge arXiv2025
La La LiDARLarge-Scale Layout Generation from LiDAR Data AAAI2026
LiDARCrafterDynamic 4D World Modeling from LiDAR Sequences AAAI2026
VeilaPanoramic LiDAR Generation from a Monocular RGB Image ICRA2026

Action Forecasters — LiDAR-Based Future Prediction 5+ papers

Model/MethodFull TitleVenueYear
Copilot4DLearning Unsupervised World Models for AD via Discrete Diffusion ICLR2024
ViDARVisual Point Cloud Forecasting enables Scalable Autonomous Driving CVPR2024
BEVWorldMultimodal World Simulator for AD via Scene-Level BEV Latents arXiv2024
HERMESA Unified Self-Driving World Model for 3D Understanding and Generation ICCV2025
DriveXOmni Scene Modeling for Generalizable World Knowledge arXiv2025

Autoregressive Simulators — Sequential LiDAR World Generation 5+ papers

Model/MethodFull TitleVenueYear
HoloDriveHolistic 2D-3D Multi-Modal Street Scene Generation for AD arXiv2024
LidarDMGenerative LiDAR Simulation in a Generated World ICRA2025
OpenDWMOpen Driving World Models arXiv2025
LiDARCrafterDynamic 4D World Modeling from LiDAR Sequences AAAI2026