BARK Simulation Dataset

该数据集是通过BARK模拟框架生成的，专注于一种场景：一辆自主车辆汇入高速公路。在此场景中，其他车辆初始条件和行为参数根据预定义的分布进行采样。数据集包含了一系列的模拟片段，这些片段在达到目标、发生碰撞或达到最大步数时结束。同时，该数据集采用了反事实策略评估方法来评价自主车辆的行为。任务涉及的是汇入车道策略评估及强化学习。

This dataset is generated via the BARK simulation framework, focusing on the scenario of an autonomous vehicle merging onto a highway. In this scenario, the initial conditions and behavioral parameters of other vehicles are sampled according to predefined distributions. The dataset consists of a series of simulation segments, which terminate when reaching the target, undergoing a collision, or hitting the maximum step count. Meanwhile, this dataset adopts counterfactual policy evaluation methods to assess the behavior of autonomous vehicles. The task involves lane-merging policy evaluation and reinforcement learning.