Hydra-MDP++: Advancing End-to-End Driving via Hydra-Distillation with Expert-Guided Decision Analysis

26 Sept 2024 (modified: 13 Nov 2024)ICLR 2025 Conference Withdrawn SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Keywords: end-to-end autonomous driving, expert guidance, knowledge distillation, open-loop metrics
Abstract: We introduce HydraMDP++, a novel end-to-end autonomous driving framework that integrates rule-based and neural planners by learning from human demonstrations and distilling knowledge from rule-based experts. We propose a teacher-student knowledge distillation framework with a multi-head student decoder that integrates feedback from rule-based expert teachers. The student model achieves state-of-the-art performance on the NAVSIM benchmark with a tiny image encoder. Moreover, to address limitations in existing evaluation metrics, we expand the teacher model to include traffic light compliance, lane-keeping ability, and extended comfort. This is intended to ensure a more robust decision synthesis in driving. HydraMDP++ demonstrates robust and efficient performance across diverse driving scenarios, achieving a 91.0% drive score on NAVSIM by simply scaling the image encoder. Our work contributes to developing more reliable and adaptable autonomous driving systems that combine the strengths of rule-based and neural planning approaches.
Primary Area: applications to robotics, autonomy, planning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 6004
Loading