CARLA Town05 Dataset

该数据集由在CARLA模拟器的Town05环境中收集的样本组成，涵盖了各种驾驶场景、控制输出和环境条件。此外，数据集还包括了多种天气状况和对抗性驾驶场景，以提高训练的鲁棒性。初始样本量约为22,000个，这些样本来自基于规则的自动驾驶演示，并通过DAgger方法收集了更多样本。该数据集的任务是训练和评估强化学习代理，以修正自动驾驶中的故障场景。

This dataset comprises samples collected in the Town05 environment of the CARLA simulator, covering diverse driving scenarios, control outputs and environmental conditions. Furthermore, it includes various weather conditions and adversarial driving scenarios to enhance the robustness of training. The initial sample size is approximately 22,000, with these samples sourced from rule-based autonomous driving demonstrations, and additional samples were collected via the DAgger method. The task of this dataset is to train and evaluate reinforcement learning agents for rectifying fault scenarios in autonomous driving.

搜集汇总

背景与挑战

背景概述

CARLA Town05 Dataset是一个专注于自动驾驶研究的数据集，特别用于模仿学习和强化学习算法的训练与评估。它包含多种复杂的城市驾驶场景，如动态车辆碰撞和行人碰撞，旨在提高自动驾驶代理在复杂环境中的性能。