Driving SMARTS Competition at NeurIPS 2022: Insights and Outcome

Amir Rasouli; Soheil Alizadeh; Iuliia Kotseruba; Yi Ma; Hebin Liang; Yuan Tian; Zhiyu Huang; Haochen Liu; Jingda Wu; Randy Goebel; Tianpei Yang; Matthew E. Taylor; Liam Paull; Xi Chen

Driving SMARTS Competition at NeurIPS 2022: Insights and Outcome

Amir Rasouli^*, Soheil Alizadeh^*, Iuliia Kotseruba, Yi Ma, Hebin Liang, Yuan Tian, Zhiyu Huang, Haochen Liu, Jingda Wu, Randy Goebel, Tianpei Yang, Matthew E. Taylor, Liam Paull, Xi Chen^*

^*Corresponding author for this work

Research output: Contribution to journal › Conference article › peer-review

Abstract

The Driving SMARTS (Scalable Multi-Agent Reinforcement Learning Training School) competition was designed to address one of the major challenges for autonomous driving (AD), namely adaptation to distribution shift between data used for training and inference and the problems caused by this shift in real-world conditions. The two key features of the competition are 1) a two-track structure to encourage and support a variety of approaches to solving the problem, such as reinforcement learning, offline learning, and other machine learning methods; and 2) curated data for driving scenarios of varying difficulty levels, from cruising to unprotected turns at unsignalized intersections. The competition attracted 87 participants in 53 teams. Top-ranking teams contributed a diverse set of solutions highlighting the effectiveness of different methodologies on safe motion planning for AD. This paper provides an overview of the Driving SMARTS competition, discusses its organisational and design aspects, and presents the results, insights, and promising directions for future research.

Original language	English
Pages (from-to)	73-84
Number of pages	12
Journal	Proceedings of Machine Learning Research
Volume	220
Publication status	Published - 2023
Externally published	Yes
Event	36th Annual Conference on Neural Information Processing Systems - Competition Track, NeurIPS 2022 - Virtual, Online, United States Duration: 28 Nov 2022 → 9 Dec 2022

Keywords

Autonomous Driving
Distribution Shift
Motion Planning
Offline Learning
Reinforcement Learning

Cite this

@article{dc915d9f680e4a43b8781bdf90a5c597,

title = "Driving SMARTS Competition at NeurIPS 2022: Insights and Outcome",

abstract = "The Driving SMARTS (Scalable Multi-Agent Reinforcement Learning Training School) competition was designed to address one of the major challenges for autonomous driving (AD), namely adaptation to distribution shift between data used for training and inference and the problems caused by this shift in real-world conditions. The two key features of the competition are 1) a two-track structure to encourage and support a variety of approaches to solving the problem, such as reinforcement learning, offline learning, and other machine learning methods; and 2) curated data for driving scenarios of varying difficulty levels, from cruising to unprotected turns at unsignalized intersections. The competition attracted 87 participants in 53 teams. Top-ranking teams contributed a diverse set of solutions highlighting the effectiveness of different methodologies on safe motion planning for AD. This paper provides an overview of the Driving SMARTS competition, discusses its organisational and design aspects, and presents the results, insights, and promising directions for future research.",

keywords = "Autonomous Driving, Distribution Shift, Motion Planning, Offline Learning, Reinforcement Learning",

author = "Amir Rasouli and Soheil Alizadeh and Iuliia Kotseruba and Yi Ma and Hebin Liang and Yuan Tian and Zhiyu Huang and Haochen Liu and Jingda Wu and Randy Goebel and Tianpei Yang and Taylor, \{Matthew E.\} and Liam Paull and Xi Chen",

note = "Publisher Copyright: {\textcopyright} 2023 A. Rasouli et al.; 36th Annual Conference on Neural Information Processing Systems - Competition Track, NeurIPS 2022 ; Conference date: 28-11-2022 Through 09-12-2022",

year = "2023",

language = "English",

volume = "220",

pages = "73--84",

journal = "Proceedings of Machine Learning Research",

issn = "2640-3498",

publisher = "ML Research Press",

}

TY - JOUR

T1 - Driving SMARTS Competition at NeurIPS 2022

T2 - 36th Annual Conference on Neural Information Processing Systems - Competition Track, NeurIPS 2022

AU - Rasouli, Amir

AU - Alizadeh, Soheil

AU - Kotseruba, Iuliia

AU - Ma, Yi

AU - Liang, Hebin

AU - Tian, Yuan

AU - Huang, Zhiyu

AU - Liu, Haochen

AU - Wu, Jingda

AU - Goebel, Randy

AU - Yang, Tianpei

AU - Taylor, Matthew E.

AU - Paull, Liam

AU - Chen, Xi

PY - 2023

Y1 - 2023

N2 - The Driving SMARTS (Scalable Multi-Agent Reinforcement Learning Training School) competition was designed to address one of the major challenges for autonomous driving (AD), namely adaptation to distribution shift between data used for training and inference and the problems caused by this shift in real-world conditions. The two key features of the competition are 1) a two-track structure to encourage and support a variety of approaches to solving the problem, such as reinforcement learning, offline learning, and other machine learning methods; and 2) curated data for driving scenarios of varying difficulty levels, from cruising to unprotected turns at unsignalized intersections. The competition attracted 87 participants in 53 teams. Top-ranking teams contributed a diverse set of solutions highlighting the effectiveness of different methodologies on safe motion planning for AD. This paper provides an overview of the Driving SMARTS competition, discusses its organisational and design aspects, and presents the results, insights, and promising directions for future research.

AB - The Driving SMARTS (Scalable Multi-Agent Reinforcement Learning Training School) competition was designed to address one of the major challenges for autonomous driving (AD), namely adaptation to distribution shift between data used for training and inference and the problems caused by this shift in real-world conditions. The two key features of the competition are 1) a two-track structure to encourage and support a variety of approaches to solving the problem, such as reinforcement learning, offline learning, and other machine learning methods; and 2) curated data for driving scenarios of varying difficulty levels, from cruising to unprotected turns at unsignalized intersections. The competition attracted 87 participants in 53 teams. Top-ranking teams contributed a diverse set of solutions highlighting the effectiveness of different methodologies on safe motion planning for AD. This paper provides an overview of the Driving SMARTS competition, discusses its organisational and design aspects, and presents the results, insights, and promising directions for future research.

KW - Autonomous Driving

KW - Distribution Shift

KW - Motion Planning

KW - Offline Learning

KW - Reinforcement Learning

UR - https://www.scopus.com/pages/publications/85179132438

M3 - Conference article

AN - SCOPUS:85179132438

SN - 2640-3498

VL - 220

SP - 73

EP - 84

JO - Proceedings of Machine Learning Research

JF - Proceedings of Machine Learning Research

Y2 - 28 November 2022 through 9 December 2022

ER -

Driving SMARTS Competition at NeurIPS 2022: Insights and Outcome

Abstract

Keywords

Other files and links

Fingerprint

Cite this