Abstract

The realization of robust and generalizable autonomous driving (AD) systems necessitates intelligent, human-like, and demonstrably safe decision-making behavior. Traditional methods, facing limitations in complex and novel scenarios, have been supplanted by modern end-to-end systems powered by imitation from vast amounts of expert data. Recent research however, highlights the surprising effectiveness of (self-play) reinforcement learning (RL), either in isolation or synergy with imitation learning (IL). This approach enables discovery of diverse skills and robust performance in out-of-distribution settings without relying purely on data and imitation of the expert as in conventional end-to-end systems. Furthermore, generative world models and Vision-Language-Action (VLA) models are revolutionizing the creation of closed-loop simulation environments by allowing for controllable and realistic scenario generation. These advancements, coupled with scalable, GPU-accelerated multi-agent training, facilitate efficient sim-to-real transfer, paving the way for safer and more capable autonomous vehicles. This workshop brings key researchers together to highlight and discuss these critical developments, with the goal of challenging assumptions and advancing understanding toward the mission of achieving robust autonomy.

Call for Papers

We welcome paper submissions on novel and interesting approaches in the field of emerging behaviors for robust autonomy, ranging from self-play RL and imitation learning to generative world models and VLA-based simulation. All papers will be carefully peer-reviewed based on their originality, relevance to the workshop topics, contributions, and technical clarity to ensure the high quality of the presented content. Accepted papers will be presented as posters. The poster session will start with a round of spotlight talks, where each author is given the opportunity to pitch their work in a 2-minute presentation. Every paper has to be covered in-person by one of the authors. Additionally, two selected PhD student papers will be given a 15-minute oral presentation during the main program.

Topics of Interest:

  • Advances in Self-Play and Multi-Agent Reinforcement Learning for driving.
  • Synergies between Imitation Learning, Offline RL, and online fine-tuning for behavior generation.
  • Methods for discovering emergent skills and complex, human-like behaviors without dense reward engineering.
  • Vision-Language-Action models and RL for end-to-end, vision-centric driving policies.
  • GPU-enabled simulation of traffic agents. Generative world models for creating diverse and controllable scenarios.
  • The intersection of RL for behavior and RL for reasoning.
  • Techniques for grounding self-play policies with real-world data and ensuring physical plausibility.
  • Analysis of failure modes, robustness, and current limitations of learned policies when deployed.
  • Integration of VLAs with fast simulators.
  • Challenges and solutions for integrating large models with real-time simulators.

Important Dates:

  • Deadline for Paper submissions: TBA, 2026
  • Notification of acceptance: TBA, 2026

Submission Instructions:

  • Papers must not exceed TBA pages (excluding references and appendices)
  • Please refer to the main conference’s format guidelines and template for detailed instructions
  • Submissions are handled via CMT. You can submit your paper here: TBA

Tentative program

Time Speaker Topic
9:00 - 9:15 Organizers Welcome and Introduction
9:15 - 09:45
Eugene Vinitsky, New York University TBA TBA
09:45 - 10:15
Abhinav Valada, University of Freiburg TBA TBA
10:15 - 10:30 Coffee Break
10:30 - 11:00
Rowan McAllister, Toyota TBA TBA
11:00 - 11:30
PhD Students Session
11:30 - 12:00
Vassia Simaiaki, Wayve TBA TBA
12:00 - 13:00 Lunch
13:00 - 14:00
Poster Session Posters: TBA
14:00 - 14:30
Max Igl, Nvidia TBA TBA
14:30 - 15:00
Bernhard Jaeger, University of Tübingen TBA TBA
15:00 - 15:15 Coffee Break
15:15 - 15:45
Hongyang Li, University of Hong Kong TBA TBA
15:45 - 16:45
Kate Tolstaya, Max Jiang, Sergio Casas, Waymo TBA TBA
16:45 - 17:00 Organizers Closing

Organizers

Luigi Palmieri,
Robert Bosch GmbH
Boris Ivanovic,
NVIDIA
Faris Janjos,
Robert Bosch GmbH
Andreas Look,
Coburg University of Applied Sciences
Daphne Cornelisse,
New York University
Shuhan Tan,
University of Texas at Austin
Max Naumann,
Zoox

Supported by