Safe Learning for Autonomous Driving

ICML 2022 Workshop + Challenge


Welcome to the 1st Workshop on Safe Learning for Autonomous Driving (SL4AD), co-located with the International Conference on Machine Learning (ICML 2022), to be held on 22 July 2022 in hybrid format.

While there have been significant advances in vehicle autonomy (e.g., perception, trajectory forecasting, planning and control, etc.), it is of paramount importance for autonomous systems to adhere to safety specifications, as any safety infraction in urban and highway driving, or high-speed racing could lead to catastrophic failures. Given this inherent tension between safety and performance, we introduce a new simulation environment in autonomous racing as a particularly challenging proving ground for safe learning algorithms.

We envision this workshop bringing together researchers and industry practitioners from different AI subfields to work towards safer and more robust autonomous technology. We encourage participants to take part in the Challenge by competing for top leaderboard positions, to submit articles for review, and to engage with us at ICML 2022.

Topics covered

  • Safe reinforcement learning, safe exploration, constrained reinforcement learning, safe learning + control theory
  • Safety verification, certifying learning-based control under dynamical uncertainty, dependability analysis
  • Robustness to out-of-distribution road scenes
  • Learning vehicle dynamics at high-speeds and in unstable regimes
  • Vision-based perception and scene understanding for autonomous driving
  • Representation learning for visuomotor control
  • Transfer learning; simulation to real-world; meta-learning; domain adaptation; *-shot learning; self/semi/weakly-supervised learning; multi-task learning
  • End-to-end and real-time autonomous driving systems
  • Novel automotive sensors and their applications
  • Trajectory forecasting; Behavior prediction of pedestrians, vehicles, and animals
  • Explainability in autonomous driving
  • Learning to drive via imitation learning
  • Learning to drive via distribution awareness
  • Uncertainty propagation through autonomous driving pipelines
  • Classical planning and control for autonomous driving
  • Cooperative and competitive multi-agent systems
  • Visual grounding and its application to autonomous driving
  • Vision-language navigation for autonomous driving
  • Audio-visual navigation for autonomous driving
  • Neuro-symbolic approaches in autonomous driving; Knowledge representation and reasoning
  • Auditory perception (detection, tracking, segmentation, motion estimation, etc)
  • Brain-inspired autonomous control systems
  • Human factors in autonomous driving
  • AI ethics in autonomous driving
  • Autonomous driving datasets, simulation, evaluations, and metrics
  • Connected autonomous driving, vehicle-to-vehicle, vehicle-to-infrastructure communication, digital twins
  • Autonomous driving for traffic management and emission reduction; intelligent transportation systems


Note: all deadlines are in Anywhere on Earth.

Paper Submission

Submissions open: 23 March 2022
Submissions due:

20 May 2022

 27 May 2022

Notification: 6 June 2022
Camera Ready: 17 June 2022
Oral/Poster video upload: 1 July 2022

Workshop Event

Date: 22 July 2022


Friday, 22 July, 2022. All times are in Eastern Daylight Time (EDT). Current time is .

Opening Remarks
Melanie Zeilinger

Melanie Zeilinger

ETH Zürich

Learning for High Performance yet Safe Control
Social + Poster Session
Gathertown and in-person displays
Spotlight Talks

Zijian Guo


#16: Constrained Model-based Reinforcement Learning via Robust Planning

Linrui Zhang

Tsinghua University

#12: SafeRL-Kit: Evaluating Efficient Reinforcement Learning Methods for Safe Autonomous Driving
Autonomous Racing Virtual Challenge: Contributed Talks

Shivansh Beohar

IIIT Allahabad

#11: Solving Learn-to-Race Autonomous Racing Challenge by Planning in Latent Space

James Bockman

U Adelaide

#14: The Edge of Disaster: A Battle Between Autonomous Racing and Safety
Lunch Break
Peter Stone

Peter Stone

UT Austin; Sony AI

Reward (Mis)design for Autonomous Driving and Accumulating Safety Rules from Catastrophic Action Effects
Spotlight Talks

Weiran Yao


#23: Distribution-aware Goal Prediction and Conformant Model-based Planning for Safe Autonomous Driving

Zuxin Liu


#15: On the Robustness of Safe Reinforcement Learning under Observational Perturbations
Todd Hester

Todd Hester

Amazon Scout

Multi-modal sensor fusion for the Amazon Scout robot
Chelsea Finn

Chelsea Finn

Stanford + Google Brain

Learning How To Be Safe
Social + Poster Session
Gathertown and in-person displays
Andrea Bajcsy

Andrea Bajcsy

UC Berkeley / CMU

Practical Safety Assurances for Dynamic Human-Robot Interactions
Jeff Schneider

Jeff Schneider

Carnegie Mellon University

Reinforcement Learning for self-driving cars
Sergey Levine

Sergey Levine

UC Berkeley

Learning Plannable Representations and Planning with Learnable Skills
Closing Remarks



We also feature an exciting and new AI Challenge in high-speed autonomous racing. Here, the goal is to evaluate the joint safety, performance, and generalisation capabilities of perception and control algorithms, as they operate simulated Formula-style racing vehicles at their physical limits! The Learn-to-Race Autonomous Racing Virtual Challenge is now active. Participate now!

L2R Autonomous Racing Virtual Challenge: Safe Learning for Autonomous Driving
L2R Autonomous Racing Virtual Challenge: Safe Learning for Autonomous Driving
L2R Autonomous Racing Virtual Challenge: Safe Learning for Autonomous Driving
L2R Autonomous Racing Virtual Challenge: Safe Learning for Autonomous Driving

Steps to victory!


Jonathan Francis

Jonathan Francis

Research Scientist at Bosch Research, focusing domain knowledge-enhanced representation learning, applied to robotics and autonomous driving

Hitesh Arora

Hitesh Arora

Researcher at Amazon, focusing on multimodal perception and reinforcement learning, applied to autonomous driving

Bingqing Chen

Bingqing Chen

Machine Learning Research Scientist at Bosch Research, focusing on constraint-based optimisation, physical mechanisms, and safe learning, applied to autonomous driving

Xinshuo Weng

Xinshuo Weng

Research Scientist at NVIDIA Research; focusing on 3D computer vision and generative models for autonomous systems

Siddha Ganju

Siddha Ganju

Researcher and Data Scientist at NVIDIA, focusing on computer vision optimization for vehicle autonomy and medical instruments

Manoj Bhat

Manoj Bhat

Machine Learning Researcher at Amazon, focusing on Neural Radiance fields, reinforcement learning, and representation learning applied to autonomous driving

Daniel Omeiza

Daniel Omeiza

PhD student at Oxford, focusing on explainable AI and decision-making, in autonomous driving

Jean Oh

Jean Oh

Research Professor in Robotics Institute at CMU and Director of Bot Intelligence Group; multimodal perception, navigation, and artificial intelligence

Eric Nyberg

Eric Nyberg

Professor of Computer Science at CMU and Program Director, Masters of Computational Data Science; hybrid reasoning systems and artificial intelligence

Sylvia Herbert

Sylvia Herbert

Assistant Professor at UCSD and Director of Safe Autonomous Systems Lab; uncertainty modeling in control, safety-aware learning, autonomy

Li Erran Li

Li Erran Li

Head of Science/HIL at Amazon AWS AI and Adjunct Professor at Columbia; Computer Vision, Machine Learning Systems, and Algorithms

Program Committee

  • Arav Agarwal (ER, TR)
  • Eren Aksoy
  • Raghuram Mandyam Annasamy
  • Tiago Cortinhal
  • Xiaoxiao Du (TR)
  • Isht Dwivedi
  • Hesham Eraqi
  • Xiangyu Gao
  • Sahika Genc (ER)
  • S. Alireza Golestaneh (TR)
  • David Held (SMR)
  • Todd Hester (SMR)
  • Zehao Huang
  • Fabian Hüger
  • Arec Jamgochian (TR)
  • Rowan McAllister (SMR)
  • Kunal Menda
  • Aarati Noronha
  • Praveen Palanisamy
  • João Pinho (ER)
  • Daniele Reda
  • Nazmus Sakib
  • Pranjay Shyam
  • Mark Schutera
  • Zhaoen Su
  • Ram Vasudevan
  • Yujie Wei
  • Weiran Yao

ERRecognises PC member who served (“+” additionally) as an Emergency Reviewer.
TRRecognises PC member who, according to Chair ratings, ranked in the Top 15% of Reviewers.
SMRRecognises PC member who agreed to provide their services as a Senior Meta-Reviewer.