Autonomous Vehicle Failure Scenarios

该数据集记录了自动驾驶车辆与行人交互的动作序列，特别关注了通过应用RSS奖励增强和通用AST生成的碰撞场景。该数据集包含了多种失败类型和轨迹，突显了奖励增强在发现多样化失败模式方面的有效性。在实验中，共记录了1000个动作序列，其任务是对自动驾驶车辆的失效检测和场景生成进行验证。

This dataset documents action sequences of interactions between autonomous vehicles and pedestrians, with a specific focus on collision scenarios generated through RSS reward enhancement and general AST. It includes a wide range of failure types and trajectories, underscoring the efficacy of reward enhancement in identifying diverse failure patterns. In total, 1000 action sequences were recorded in the experiments, which are employed to validate failure detection for autonomous vehicles and scenario generation.