CAT: Closed-loop Adversarial Training for Safe End-to-End Driving
CAT: Closed-loop Adversarial Training for Safe End-to-End Driving
复制标题
DOI：
10.48550/arxiv.2310.12432
复制
发表时间：
2023-10
期刊：
影响因子：
--
通讯作者：
Linrui Zhang;Zhenghao Peng;Quanyi Li;Bolei Zhou
中科院分区：
文献类型：
--
作者：
Linrui Zhang;Zhenghao Peng;Quanyi Li;Bolei Zhou
驾驶安全是自动驾驶汽车的首要任务。与以往处理事故多发交通事件的算法设计在策略层面上的工作正交，本文从环境增强的角度研究了一种端到端安全驾驶的闭环对抗性训练(CAT)框架。CAT旨在通过对驾驶代理进行安全关键场景培训，不断提高驾驶代理的安全性，这些场景是随着时间的推移动态生成的。提出了一种新的重采样技术，通过概率因式分解将日志重放的真实驾驶场景转化为安全关键场景，其中对抗性流量的生成被建模为标准运动预测子问题的乘法。因此，与现有的安全关键场景生成方法相比，CAT可以发起更有效的物理攻击，并在迭代学习管道中产生显著更低的计算成本。我们将CAT集成到MetaDrive模拟器中，并在从真实驾驶数据集导入的数百个驾驶场景上验证了我们的方法。实验结果表明，CAT能够有效地生成对抗被训练智能体的对抗性场景。经过培训后，该代理可以在日志重放和安全关键交通场景中在坚持测试集上实现卓越的驾驶安全。有关代码和数据，请访问https://metadriverse.github.io/cat.
Driving safety is a top priority for autonomous vehicles. Orthogonal to prior work handling accident-prone traffic events by algorithm designs at the policy level, we investigate a Closed-loop Adversarial Training (CAT) framework for safe end-to-end driving in this paper through the lens of environment augmentation. CAT aims to continuously improve the safety of driving agents by training the agent on safety-critical scenarios that are dynamically generated over time. A novel resampling technique is developed to turn log-replay real-world driving scenarios into safety-critical ones via probabilistic factorization, where the adversarial traffic generation is modeled as the multiplication of standard motion prediction sub-problems. Consequently, CAT can launch more efficient physical attacks compared to existing safety-critical scenario generation methods and yields a significantly less computational cost in the iterative learning pipeline. We incorporate CAT into the MetaDrive simulator and validate our approach on hundreds of driving scenarios imported from real-world driving datasets. Experimental results demonstrate that CAT can effectively generate adversarial scenarios countering the agent being trained. After training, the agent can achieve superior driving safety in both log-replay and safety-critical traffic scenarios on the held-out test set. Code and data are available at https://metadriverse.github.io/cat.