SafeBench
收藏资源简介:
SafeBench是由伊利诺伊大学厄巴纳-香槟分校和卡内基梅隆大学共同开发的自动驾驶安全评估基准平台。该数据集包含2352个安全关键测试场景,涵盖了国家公路交通安全管理局定义的8种安全关键测试场景,并针对每种场景设计了10种不同的变体。SafeBench通过集成多种场景生成算法和驾驶路线变体,旨在为自动驾驶算法提供一个统一的平台,以评估其在各种环境下的性能。该数据集的应用领域主要集中在自动驾驶系统的安全性和鲁棒性评估,旨在解决自动驾驶系统在复杂和罕见情况下的性能问题。
SafeBench is an autonomous driving safety evaluation benchmark platform co-developed by the University of Illinois Urbana-Champaign and Carnegie Mellon University. This dataset comprises 2352 safety-critical test scenarios, covering 8 categories of safety-critical test scenarios defined by the National Highway Traffic Safety Administration (NHTSA), with 10 distinct variants developed for each scenario. By integrating multiple scenario generation algorithms and driving route variants, SafeBench aims to provide a unified platform for autonomous driving algorithms to evaluate their performance across diverse environments. The dataset is primarily applied to the safety and robustness evaluation of autonomous driving systems, aiming to address the performance issues of autonomous driving systems in complex and rare scenarios.




