CarDreamer Driving Tasks

该数据集是一套与Gym接口兼容的、全面的、可配置的驾驶任务集合，旨在评估世界模型在自动驾驶场景中的表现。它包含了多种评价指标，如成功率、平均距离、碰撞率、平均速度以及航点距离等，以全面评估代理在不同驾驶条件下的性能。此外，该数据集还采用了小型DreamerV3模型，拥有1800万参数，并在单个NVIDIA 4090 GPU上进行训练，其任务是评估自动驾驶任务的性能。

This dataset is a comprehensive, configurable collection of driving tasks compatible with the Gym interface, designed to evaluate the performance of world models in autonomous driving scenarios. It encompasses a variety of evaluation metrics, including success rate, average travel distance, collision rate, average speed, waypoint distance and others, to comprehensively assess the performance of agents across diverse driving conditions. Additionally, this dataset integrates a small-scale DreamerV3 model with 18 million parameters, which is trained on a single NVIDIA 4090 GPU and serves to evaluate the performance of autonomous driving tasks.