arXiv 2409.129972024-09-19cs.LGcs.AI

VCAT：用于增强自动驾驶车辆鲁棒性的漏洞感知与好奇心驱动的对抗训练

VCAT: Vulnerability-aware and Curiosity-driven Adversarial Training for Enhancing Autonomous Vehicle Robustness

State Key Laboratory of Intelligent Transportation Systems（智能交通系统国家重点实验室）；School of Transportation Science and Technology, Beihang University（北京航空航天大学交通科学与工程学院）；Department of Civil and Environmental Engineering, Rensselaer Polytechnic Institute（伦斯勒理工学院土木与环境工程系）

Xuan Cai, Zhiyong Cui, Xuesong Bai, Ruimin Ke, Zhenshu Ma, Haiyang Yu, Yilong Ren

备注：: 7 pages, 5 figures, conference

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对自动驾驶车辆对抗训练中攻击者易陷入过度利用已知漏洞的问题，提出 VCAT 框架，利用代理网络感知受害者漏洞并结合随机网络蒸馏构建内在奖励引导探索，显著提升了自动驾驶车辆的鲁棒性并降低了碰撞率。

摘要：

自动驾驶车辆（AVs）在复杂交通环境中的安全运行面临重大威胁。对抗训练已成为一种有效方法，使自动驾驶车辆能够预先增强其抵御恶意攻击的鲁棒性。该方法通过使用对抗策略训练攻击者，允许自动驾驶车辆通过与该攻击者交互来学习鲁棒驾驶。然而，现有方法中的对抗策略往往陷入过度利用已知漏洞的循环，导致自动驾驶车辆的鲁棒性提升不佳。为克服这些限制，我们引入了一个名为 Vulnerability-aware and Curiosity-driven Adversarial Training (VCAT) 的开创性框架。具体而言，在交通车辆攻击者训练阶段，采用一个代理网络来拟合自动驾驶车辆受害者的价值函数，提供关于受害者内在漏洞的密集信息。随后，使用随机网络蒸馏来表征环境的新颖性，构建内在奖励以引导攻击者探索未知领域。在受害者防御训练阶段，自动驾驶车辆在关键场景中进行训练，其中预训练的攻击者被放置在受害者周围以生成攻击行为。实验结果表明，VCAT 提供的训练方法显著提升了基于学习的自动驾驶车辆的鲁棒控制能力，优于传统训练模式和其他强化学习同类方法，并显著降低了碰撞率。代码可在 https://github.com/caixxuan/VCAT 获取。

英文摘要：

Autonomous vehicles (AVs) face significant threats to their safe operation in complex traffic environments. Adversarial training has emerged as an effective method of enabling AVs to preemptively fortify their robustness against malicious attacks. Train an attacker using an adversarial policy, allowing the AV to learn robust driving through interaction with this attacker. However, adversarial policies in existing methodologies often get stuck in a loop of overexploiting established vulnerabilities, resulting in poor improvement for AVs. To overcome the limitations, we introduce a pioneering framework termed Vulnerability-aware and Curiosity-driven Adversarial Training (VCAT). Specifically, during the traffic vehicle attacker training phase, a surrogate network is employed to fit the value function of the AV victim, providing dense information about the victim's inherent vulnerabilities. Subsequently, random network distillation is used to characterize the novelty of the environment, constructing an intrinsic reward to guide the attacker in exploring unexplored territories. In the victim defense training phase, the AV is trained in critical scenarios in which the pretrained attacker is positioned around the victim to generate attack behaviors. Experimental results revealed that the training methodology provided by VCAT significantly improved the robust control capabilities of learning-based AVs, outperforming both conventional training modalities and alternative reinforcement learning counterparts, with a marked reduction in crash rates. The code is available at https://github.com/caixxuan/VCAT.

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。