arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2507.04464cs.AI
收藏

基于逆强化学习的异常决策发现

Anomalous Decision Discovery using Inverse Reinforcement Learning

School of Computing, Clemson University(计算学院,克莱姆森大学)

Ashish Bastola, Mert D. Pesé, Long Cheng, Jonathon Smereka, Abolfazl Razi

AI总结:

本文提出基于逆强化学习的TRAP框架,通过奖励与最坏情况监督学习时间信用分配,在自动驾驶异常检测中实现噪声鲁棒和未见场景泛化,实验表现优于基线。

摘要:

异常检测在自动驾驶车辆(AVs)中发挥着关键作用,它通过感知系统识别可能危及安全并导致危险情况的异常行为。当前的方法通常依赖预定义阈值或监督学习范式,在面对未见场景、传感器噪声和遮挡时效果下降,可能导致安全关键性失效。此外,监督方法需要大规模标注数据集,限制了其在现实世界中的可行性。为解决这些不足,我们提出了一种基于逆强化学习(IRL)的异常检测框架,从序列感知数据中推断潜在驾驶意图,从而实现鲁棒识别。具体而言,我们提出了轨迹-奖励引导的自适应预训练(TRAP),这是一种用于异常检测的新型IRL框架,旨在解决现有方法的两个关键局限:噪声鲁棒性和对未见场景的泛化能力。我们的核心创新是通过奖励和最坏情况监督隐式学习时间信用分配。我们利用可变时域采样进行预训练,以最大化到后果发生的时间,从而实现行为偏差的早期检测。在14,000多条模拟轨迹上的实验展示了最先进的性能,达到0.90 AUC和82.2%的F1分数——在召回率上比同样训练的监督和无监督基线分别高出39%,在F1分数上高出12%。在表现出对各种噪声类型的鲁棒性和对未见异常类型的泛化能力的同时,也取得了相似的性能。我们的代码将发布于:https://github.com/abastola0/TRAP.git

英文摘要:

Anomaly detection plays a critical role in Autonomous Vehicles (AVs) by identifying unusual behaviors through perception systems that could compromise safety and lead to hazardous situations. Current approaches, which often rely on predefined thresholds or supervised learning paradigms, exhibit reduced efficacy when confronted with unseen scenarios, sensor noise, and occlusions, leading to potential safety-critical failures. Moreover, supervised methods require large annotated datasets, limiting their real-world feasibility. To address these gaps, we propose an anomaly detection framework based on Inverse Reinforcement Learning (IRL) to infer latent driving intentions from sequential perception data, thus enabling robust identification. Specifically, we present Trajectory-Reward Guided Adaptive Pre-training (TRAP), a novel IRL framework for anomaly detection, to address two critical limitations of existing methods: noise robustness and generalization to unseen scenarios. Our core innovation is implicitly learning temporal credit assignments via reward and worst-case supervision. We leverage pre-training with variable-horizon sampling to maximize time-to-consequence, resulting in early detection of behavior deviation. Experiments on 14,000+ simulated trajectories demonstrate state-of-the-art performance, achieving 0.90 AUC and 82.2\% F1-score - outperforming similarly trained supervised and unsupervised baselines by 39\% on Recall and 12\% on F1-score, respectively. Similar performance is achieved while exhibiting robustness to various noise types and generalization to unseen anomaly types. Our code will be available at: https://github.com/abastola0/TRAP.git

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。

↑