From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xia, Xinyu, Ma, Xingjun, Hu, Yunfeng, Qu, Ting, Chen, Hong, Gong, Xun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913863594672128
author Xia, Xinyu
Ma, Xingjun
Hu, Yunfeng
Qu, Ting
Chen, Hong
Gong, Xun
author_facet Xia, Xinyu
Ma, Xingjun
Hu, Yunfeng
Qu, Ting
Chen, Hong
Gong, Xun
contents Ensuring robust and generalizable autonomous driving requires not only broad scenario coverage but also efficient repair of failure cases, particularly those related to challenging and safety-critical scenarios. However, existing scenario generation and selection methods often lack adaptivity and semantic relevance, limiting their impact on performance improvement. In this paper, we propose \textbf{SERA}, an LLM-powered framework that enables autonomous driving systems to self-evolve by repairing failure cases through targeted scenario recommendation. By analyzing performance logs, SERA identifies failure patterns and dynamically retrieves semantically aligned scenarios from a structured bank. An LLM-based reflection mechanism further refines these recommendations to maximize relevance and diversity. The selected scenarios are used for few-shot fine-tuning, enabling targeted adaptation with minimal data. Experiments on the benchmark show that SERA consistently improves key metrics across multiple autonomous driving baselines, demonstrating its effectiveness and generalizability under safety-critical conditions.
format Preprint
id arxiv_https___arxiv_org_abs_2505_22067
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving
Xia, Xinyu
Ma, Xingjun
Hu, Yunfeng
Qu, Ting
Chen, Hong
Gong, Xun
Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
Ensuring robust and generalizable autonomous driving requires not only broad scenario coverage but also efficient repair of failure cases, particularly those related to challenging and safety-critical scenarios. However, existing scenario generation and selection methods often lack adaptivity and semantic relevance, limiting their impact on performance improvement. In this paper, we propose \textbf{SERA}, an LLM-powered framework that enables autonomous driving systems to self-evolve by repairing failure cases through targeted scenario recommendation. By analyzing performance logs, SERA identifies failure patterns and dynamically retrieves semantically aligned scenarios from a structured bank. An LLM-based reflection mechanism further refines these recommendations to maximize relevance and diversity. The selected scenarios are used for few-shot fine-tuning, enabling targeted adaptation with minimal data. Experiments on the benchmark show that SERA consistently improves key metrics across multiple autonomous driving baselines, demonstrating its effectiveness and generalizability under safety-critical conditions.
title From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2505.22067