SimScale: Learning to Drive via Real-World Simulation at Scale

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Tian, Haochen, Li, Tianyu, Liu, Haochen, Yang, Jiazhi, Qiu, Yihang, Li, Guang, Wang, Junli, Gao, Yinfeng, Zhang, Zhang, Wang, Liang, Ye, Hangjun, Tan, Tieniu, Chen, Long, Li, Hongyang
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913020359213056
author Tian, Haochen
Li, Tianyu
Liu, Haochen
Yang, Jiazhi
Qiu, Yihang
Li, Guang
Wang, Junli
Gao, Yinfeng
Zhang, Zhang
Wang, Liang
Ye, Hangjun
Tan, Tieniu
Chen, Long
Li, Hongyang
author_facet Tian, Haochen
Li, Tianyu
Liu, Haochen
Yang, Jiazhi
Qiu, Yihang
Li, Guang
Wang, Junli
Gao, Yinfeng
Zhang, Zhang
Wang, Liang
Ye, Hangjun
Tan, Tieniu
Chen, Long
Li, Hongyang
contents Achieving fully autonomous driving systems requires learning rational decisions in a wide span of scenarios, including safety-critical and out-of-distribution ones. However, such cases are underrepresented in real-world corpus collected by human experts. To complement for the lack of data diversity, we introduce a novel and scalable simulation framework capable of synthesizing massive unseen states upon existing driving logs. Our pipeline utilizes advanced neural rendering with a reactive environment to generate high-fidelity multi-view observations controlled by the perturbed ego trajectory. Furthermore, we develop a pseudo-expert trajectory generation mechanism for these newly simulated states to provide action supervision. Upon the synthesized data, we find that a simple co-training strategy on both real-world and simulated samples can lead to significant improvements in both robustness and generalization for various planning methods on challenging real-world benchmarks, up to +8.6 EPDMS on navhard and +2.9 on navtest. More importantly, such policy improvement scales smoothly by increasing simulation data only, even without extra real-world data streaming in. We further reveal several crucial findings of such a sim-real learning system, which we term SimScale, including the design of pseudo-experts and the scaling properties for different policy architectures. Simulation data and code have been released at https://github.com/OpenDriveLab/SimScale.
format Preprint
id arxiv_https___arxiv_org_abs_2511_23369
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SimScale: Learning to Drive via Real-World Simulation at Scale
Tian, Haochen
Li, Tianyu
Liu, Haochen
Yang, Jiazhi
Qiu, Yihang
Li, Guang
Wang, Junli
Gao, Yinfeng
Zhang, Zhang
Wang, Liang
Ye, Hangjun
Tan, Tieniu
Chen, Long
Li, Hongyang
Computer Vision and Pattern Recognition
Robotics
Achieving fully autonomous driving systems requires learning rational decisions in a wide span of scenarios, including safety-critical and out-of-distribution ones. However, such cases are underrepresented in real-world corpus collected by human experts. To complement for the lack of data diversity, we introduce a novel and scalable simulation framework capable of synthesizing massive unseen states upon existing driving logs. Our pipeline utilizes advanced neural rendering with a reactive environment to generate high-fidelity multi-view observations controlled by the perturbed ego trajectory. Furthermore, we develop a pseudo-expert trajectory generation mechanism for these newly simulated states to provide action supervision. Upon the synthesized data, we find that a simple co-training strategy on both real-world and simulated samples can lead to significant improvements in both robustness and generalization for various planning methods on challenging real-world benchmarks, up to +8.6 EPDMS on navhard and +2.9 on navtest. More importantly, such policy improvement scales smoothly by increasing simulation data only, even without extra real-world data streaming in. We further reveal several crucial findings of such a sim-real learning system, which we term SimScale, including the design of pseudo-experts and the scaling properties for different policy architectures. Simulation data and code have been released at https://github.com/OpenDriveLab/SimScale.
title SimScale: Learning to Drive via Real-World Simulation at Scale
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2511.23369