The Role of World Models in Shaping Autonomous Driving: A Comprehensive Survey

arXiv 25.02 2025 67.2 survey

TLDR

Comprehensive survey of Driving World Models for autonomous driving, covering simulators, datasets, modalities, and applications.

Reasoning

The paper provides a thorough categorization of DWM approaches and ecosystem components, which is a strength. However, as a survey, it lacks novel contributions or experimental results, and the depth of analysis may be limited.

Read-first score

Read-first score 67.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 52.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Topical relevance 42%

74.3

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=dataset,evaluation,metric

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset,github

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 113.

Keyword Scores

world model

video world model

world dynamics prediction

generative world model

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Comprehensive overview of the Driving World Model (DWM) ecosystem including simulators, datasets, and evaluation metrics
Categorization of DWM approaches based on predicted scene modalities: video, point cloud, occupancy, latent feature, and traffic map
Performance comparison of representative DWM approaches across generation and driving tasks

Methodology

This survey systematically reviews the DWM ecosystem by examining mainstream simulators, high-impact datasets, and multi-dimensional evaluation metrics. It categorizes existing DWM approaches by the modality of predicted scenes (video, point cloud, occupancy, latent feature, traffic map) and summarizes their applications in autonomous driving research. The performance of representative methods is presented for both generation and driving tasks.

Key Results

The survey provides a comprehensive overview and performance comparison of DWM approaches, demonstrating their effectiveness in scene prediction and driving tasks, and identifies potential limitations and future directions.

Resources

Paper