PDM-Lite Dataset for CARLA Leaderboard 2.0

Description

PDM-Lite is a state-of-the-art rule-based expert system for autonomous urban driving in CARLA Leaderboard 2.0, and the first to successfully navigate all scenarios. This dataset was used to create the QA dataset for DriveLM-Carla, a benchmark for evaluating end-to-end autonomous driving algorithms with Graph Visual Question Answering (GVQA). DriveLM introduces GVQA as a novel approach, modeling perception, prediction, and planning through interconnected question-answer pairs, mimicking human reasoning processes. Additionally, this dataset was used for training Transfuser++ with imitation learning, which achieved 1st place (map track) and 2nd place (sensor track) in the CARLA Autonomous Driving Challenge 2024. This dataset builds upon the PDM-Lite expert, incorporating enhancements from "Tackling CARLA Leaderboard 2.0 with End-to-End Imitation Learning".

For more information and a script for downloading and unpacking visit our GitHub.

Dataset Features

High-Quality Data: 5,134 routes with 100 % route completion and zero infractions on 8 towns, sampled at 4 Hz, totaling 581,662 frames
Diverse Scenarios: Covers 38 complex scenarios, including urban traffic, participants violating traffic rules, and high-speed highway driving
Focused Evaluation: Short routes averaging 160 m in length

Data Modalities

BEV Semantics Map: 512x512 pixels, centered on ego vehicle, 2 pixels per meter resolution
Image Data: 1024x512 pixels, RGB images, semantic segmentation, and depth information
Lidar Data: Detailed lidar point clouds with 600,000 points per second
Augmented Data: Augmented versions of RGB, semantic, depth, and lidar data
Simulator Data: Comprehensive information on nearby objects

License and Citation

Apache 2.0 license unless specified otherwise.

@inproceedings{sima2024drivelm,
  title={DriveLM: Driving with Graph Visual Question Answering},
  author={Chonghao Sima and Katrin Renz and Kashyap Chitta and Li Chen and Hanxue Zhang and Chengen Xie and Jens Beißwenger and Ping Luo and Andreas Geiger and Hongyang Li},
  booktitle={European Conference on Computer Vision},
  year={2024},
}
@misc{Beißwenger2024PdmLite,
  title        = {{PDM-Lite}: A Rule-Based Planner for CARLA Leaderboard 2.0},
  author       = {Bei{\ss}wenger, Jens},
  howpublished = {\url{https://github.com/OpenDriveLab/DriveLM/blob/DriveLM-CARLA/docs/report.pdf}},
  year         = {2024},
  school       = {University of Tübingen},
}