notesum.ai

Published at December 3

Preliminary Investigation into Data Scaling Laws for Imitation Learning-Based End-to-End Autonomous Driving

cs.RO

Released Date: December 3, 2024

Authors: Yupeng Zheng¹, Zhongpu Xia², Qichao Zhang, Teng Zhang, Ben Lu, Xiaochuang Huo, Chao Han, Yixian Li, Mengjie Yu, Bu Jin, Pengxuan Yang, Yuhang Zheng, Haifeng Yuan, Ke Jiang, Peng Jia, Xianpeng Lang, Dongbin Zhao

Aff.: ¹Institute of Automation, Chinese Academy of Sciences; ²Li Auto

Arxiv: http://arxiv.org/pdf/2412.02689v1

Dataset Size (Demonstrations)	Training Resource	Training Steps
10 thousand	64 A100	$6.0\times 10^{5}$
50 thousand	64 A100	$8.6\times 10^{5}$
0.7 million	320 A100	$9.3\times 10^{6}$
2 million	512 A100	$2.2\times 10^{7}$
4 million	512 A100	$3.1\times 10^{7}$