arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.09018cs.ROcs.AIcs.CVcs.LG
收藏

鲁棒性是一种函数,而不是一个数字:对基于视觉的驾驶OOD鲁棒性进行因素化的综合研究

Robustness Is a Function, Not a Number: A Factorized Comprehensive Study of OOD Robustness in Vision-Based Driving

University of Haifa(海法大学)

Amir Mallak, Alaa Maalouf

AI总结:

本文研究了基于视觉的驾驶OOD鲁棒性,通过因素分解发现ViT策略比CNN更鲁棒,季节和时间变化显著影响性能,FM特征在延迟成本下表现最佳,多环境训练能提升弱案例表现。

摘要:

在自动驾驶中,out of distribution(OOD)鲁棒性经常被简化为一个单一的数字,这隐藏了什么会破坏策略。我们沿着五个轴分解环境:场景(乡村/城市)、季节、天气、时间(白天/夜晚)和代理混合;并在受控的k-因子扰动下(k∈{0,1,2,3})测量性能。使用VISTA中的闭环控制,我们基准测试了FC、CNN和ViT策略,训练紧凑的ViT头在冻结的基础模型(FM)特征上进行训练,并在规模、多样性和时间上下文中变化ID支持。(1)ViT策略比同样规模的CNN/FC明显更具有OOD鲁棒性,且FM特征在延迟成本下实现了最先进的成功。(2)朴素的时间输入(多帧)无法超越最佳的单帧基线。(3)最大的单因素下降是乡村→城市和白天→夜晚(各约31%);代理交换约10%,中等降雨约7%;季节变化可能剧烈,将时间翻转与其他变化结合进一步降低性能。(4)FM特征策略在三个同时变化下仍保持在85%以上;非FM单帧策略在第一次变化中遭受重大打击,所有无FM模型在三次变化后均低于50%。(5)交互是非加性的:某些配对部分抵消,而季节-时间组合尤其有害。(6)在冬季/雪中训练最能抵抗单因素变化,而乡村+夏季基线提供了最佳的OOD性能。(7)扩展轨迹/视图提高了鲁棒性(从5到14轨迹增加了11.8个点),但针对恶劣条件的定向暴露可以替代规模。(8)使用多个ID环境扩大了覆盖范围并强化了弱案例(城市OOD60.6%→70.1%)并以小ID下降;单ID保持峰值性能但领域狭窄。这些结果为OOD鲁棒驾驶策略提供了可操作的设计规则。

英文摘要:

Out of distribution (OOD) robustness in autonomous driving is often reduced to a single number, hiding what breaks a policy. We decompose environments along five axes: scene (rural/urban), season, weather, time (day/night), and agent mix; and measure performance under controlled $k$-factor perturbations ($k \in \{0,1,2,3\}$). Using closed loop control in VISTA, we benchmark FC, CNN, and ViT policies, train compact ViT heads on frozen foundation-model (FM) features, and vary ID support in scale, diversity, and temporal context. (1) ViT policies are markedly more OOD-robust than comparably sized CNN/FC, and FM features yield state-of-the-art success at a latency cost. (2) Naive temporal inputs (multi-frame) do not beat the best single-frame baseline. (3) The largest single factor drops are rural $\rightarrow$ urban and day $\rightarrow$ night ($\sim 31\%$ each); actor swaps $\sim 10\%$, moderate rain $\sim 7\%$; season shifts can be drastic, and combining a time flip with other changes further degrades performance. (4) FM-feature policies stay above $85\%$ under three simultaneous changes; non-FM single-frame policies take a large first-shift hit, and all no-FM models fall below $50\%$ by three changes. (5) Interactions are non-additive: some pairings partially offset, whereas season-time combinations are especially harmful. (6) Training on winter/snow is most robust to single-factor shifts, while a rural+summer baseline gives the best overall OOD performance. (7) Scaling traces/views improves robustness ($+11.8$ points from $5$ to $14$ traces), yet targeted exposure to hard conditions can substitute for scale. (8) Using multiple ID environments broadens coverage and strengthens weak cases (urban OOD $60.6\% \rightarrow 70.1\%$) with a small ID drop; single-ID preserves peak performance but in a narrow domain. These results yield actionable design rules for OOD-robust driving policies.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑