发表机构
University of Macau; Chongqing University; The Hong Kong University of Science and Technology; Tongji University; Beihang University（澳门大学; 重庆大学; 香港科技大学; 同济大学; 北京航空航天大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本综述梳理了端到端自动驾驶从回归任务向面向规划系统的转变，沿四个维度整合方法，分析基准转变，指出需结合一致评估解读架构进展，并总结了多项开放性挑战。
AI 中文摘要
端到端自动驾驶已从相机到控制的回归任务，发展为采用结构化表示、轨迹级输出且评估协议日益贴近实际的面向规划的系统。本综述回顾了这一转变，涵盖行为克隆、条件模仿学习、特权知识蒸馏、BEV（鸟瞰图）与矢量化规划、统一感知-预测-规划架构、基于世界模型的规划器以及视觉-语言-行动系统。我们认为，现代端到端驾驶的关键区别并非是否使用中间表示，而是这些表示是否经过学习、监督和评估，以支持安全、可行且符合路线要求的规划。为梳理相关文献，我们沿输入表示、规划输出、监督信号和评估协议四个维度整合现有方法，进一步考察了基准从开环轨迹匹配向闭环仿真、非反应式真实道路评估、长尾测试及人类偏好感知指标的转变。我们的分析表明，若缺乏与基准一致的评估，架构进展难以解读，且仅基于位移的开环指标无法为安全且符合人类预期的驾驶提供充分证据。最后，我们总结了不确定性感知规划、学习者-专家不匹配、运行时安全保障、语言-行动 grounding（接地）、世界模型验证及可复现基准测试等开放性挑战。
英文摘要
End-to-end autonomous driving has evolved from camera-to-control regression toward planning-oriented systems that use structured representations, trajectory-level outputs, and increasingly realistic evaluation protocols. This survey reviews this transition across behavior cloning, conditional imitation learning, privileged distillation, BEV and vectorized planning, unified perception-prediction-planning architectures, world-model-based planners, and vision-language-action systems. We argue that the key distinction in modern end-to-end driving is not whether intermediate representations are used, but whether they are learned, supervised, and evaluated to support safe, feasible, and route-compliant planning. To organize the literature, we synthesize existing methods along four axes: input representation, planning output, supervision signal, and evaluation protocol. We further examine the benchmark shift from open-loop trajectory matching to closed-loop simulation, non-reactive real-log evaluation, long-tail testing, and human-preference-aware metrics. Our analysis highlights that architectural progress is difficult to interpret without benchmark-consistent evaluation, and that displacement-based open-loop metrics alone provide limited evidence for safe and human-aligned driving. We conclude with open challenges in uncertainty-aware planning, learner-expert mismatch, runtime safety assurance, language-action grounding, world-model validation, and reproducible benchmarking.
