⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

DriveWorld: 4D Pre-trained Scene Understanding via World Models for Autonomous Driving

CVPR 24 2024 51.1 method, application

TLDR

DriveWorld uses world models for 4D pre-training from multi-camera videos, improving multiple autonomous driving tasks via spatio-temporal representation learning.

Reasoning

The paper introduces a novel world model-based framework with a Memory State-Space Model for spatio-temporal representation learning, achieving significant improvements across various autonomous driving tasks. However, the abstract does not compare to other world model approaches or discuss limitations.

Read-first score

Read-first score 51.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 30.

Recency 8%
75.1

Uses a gentle age decay so recent papers surface without erasing older foundations. 2024

Methodology quality 25%
70

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

Topical relevance 42%
42.9

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

Reproducibility 25%
38

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

Field roles

Methodology anchor

Rank sensitivity

Stability: volatile; rank range: 293.

Keyword Scores

world model
10
world dynamics prediction
7
video world model
5
generative world model
3
world simulator
2
interactive world model
2
model-based reinforcement learning world model
1

Deep Analysis

Innovations

  • World model-based 4D representation learning framework for autonomous driving
  • Memory State-Space Model with Dynamic Memory Bank for temporal-aware latent dynamics and Static Scene Propagation for spatial-aware latent statics
  • Task Prompt for decoupling task-aware features for various downstream tasks

Methodology

DriveWorld pre-trains from multi-camera driving videos using a world model approach. It employs a Memory State-Space Model with a Dynamic Memory Bank for temporal-aware latent dynamics and a Static Scene Propagation module for spatial-aware latent statics. A Task Prompt is introduced to decouple task-aware features for downstream tasks.

Key Results

When pre-trained on OpenScene, DriveWorld achieves a 7.5% increase in mAP for 3D object detection, a 3.0% increase in IoU for online mapping, a 5.0% increase in AMOTA for multi-object tracking, a 0.1m decrease in minADE for motion forecasting, a 3.0% increase in IoU for occupancy prediction, and a 0.34m reduction in average L2 error for planning.

Resources

Paper

Tags