- About
Autonomous vehicles (AVs) have long been regarded as a future product that can improve traffic efficiency, contribute to environmental protection, and reduce road accidents. While we see sufficient maturity and improvement recently in the self-driving stack, the large-scale deployment of fully self-driving cars is still at its early stage. In the past years, ML-first solutions have been showing promising results scaling with the amount of data, not only in simulation but also in real-world, complex environments. We want to open up the discussion on the challenges that need to be solved in order to enable the scaling of AVs in the real world and to encourage new ideas regarding their interpretability and safety.

- Paper Topics
To promote a diversity of content, any topics related to the scalability of robotics system and autonomous driving are encouraged.

A non-exhaustive list of relevant topics:

Modeling: designing of the self-driving stacks; framing the driving problem as Behaviour Cloning, Reinforcement Learning, Path Searching, or Unsupervised Learning problems

Interpretability: explainable driving decisions, methods for debugging errors

Curriculum design: design of data curriculum, collecting strategy, balancing, continuous learning methods, etc.

Generalisation: scaling the self-driving algorithm to different geo-locations and scenarios without the need for pre-mapping, re-training, etc.

Simulation: using simulators to empower and evaluate real-world driving performance, transfer learning from sim to real

Reward design: designing of human-like, efficient driving rewards

Safety: scaling the autonomous driving with safety-first principle, robustness to out-of-distribution scenarios

Evaluation: on-road or off-road virtual evaluation of the self-driving stack, design of metrics

Transition: integration and coordination with human-driven vehicles, coordination with other AVs (V2V) or infrastructure (V2I)





- Format Guidelines
Submission Portal: https://cmt3.research.microsoft.com/ScalableAD2023

Format:

Manuscripts should follow the ICRA 2023 paper template.

Direct links to ieeeconf.zip and IEEEtranBST.zip.

Paper Length: The page limit is 6 pages (excluding references and appendix).

Dual Submission: We accept dual submissions, but the manuscript must contain substantial original contents not submitted to prior conferences, workshops, or journals.

Non-archival: The workshop is a non-archival venue and will not have official proceedings. Workshop submissions can be subsequently or concurrently submitted to other venues.

Visibility: Submissions and reviews will be private. Only accepted papers will be made public.





- Review
The review process will be double-blind. As an author, you are responsible for anonymizing your submission. You should not include author names, author affiliations, or acknowledgements in your submission.

All accepted papers will be presented as posters. The guidelines for the posters are the same as at the main conference.

At least one co-author of each accepted paper is expected to register for ICRA 2023 and attend the poster session. Remote attendance permitted.

If you need a Visa to enter the US, please apply for one as early as possible. See Visa information provided on the ICRA website.

All the accepted submissions will be available on our workshop website, though authors could indicate explicitly if they want to opt out.




- Contact
icra2023-av-workshop@googlegroups.com