MetaDrive

该数据集名为MetaDrive，旨在为训练智能体完成自动驾驶任务提供一个环境。它主要关注于通过低级别的连续控制动作，操纵目标车辆到达目的地。在数据集中，训练集展示了温和的交通条件，而测试集则包含了更高的交通密度和障碍物。该数据集的规模包括50个训练环境和20个未见过的测试环境，其核心任务为自动驾驶。This dataset, named MetaDrive, is designed to provide an environment for training AI Agents to complete autonomous driving tasks. It primarily focuses on manipulating the target vehicle to reach its destination via low-level continuous control actions. In the dataset, the training set features mild traffic conditions, while the test set contains higher traffic density and obstacles. The dataset includes 50 training environments and 20 unseen test environments, with autonomous driving as its core task.

搜集汇总

背景与挑战

背景概述

MetaDrive是一个自动驾驶训练环境数据集，通过低级别连续控制动作操纵车辆到达目的地。其训练集为温和交通条件，测试集则包含更高交通密度和障碍物，共提供50个训练环境和20个未见过的测试环境，核心任务为自动驾驶。