WOMD dataset
收藏资源简介:
WOMD数据集是由Waymax模拟器基于的真实世界人类驾驶数据集,包含超过570小时的驾驶数据,覆盖了美国六个城市的1,750公里道路。数据集记录了车辆、其他道路使用者和道路特征的轨迹,采样频率为10Hz,每个场景分为91个时间步长。数据集的创建旨在研究自动驾驶中的模仿学习问题,特别是模仿差距对学习驾驶策略的影响。通过引入部分可观测性约束,数据集帮助评估和改进从人类驾驶演示中学习自动驾驶策略的方法。
The WOMD dataset is a real-world human driving dataset constructed using the Waymax simulator. It contains over 570 hours of driving data, covering 1,750 kilometers of road across six cities in the United States. The dataset records the trajectories of vehicles, other road users and road features, with a sampling frequency of 10 Hz, and each scenario is divided into 91 time steps. The dataset was created to study imitation learning problems in autonomous driving, particularly the impact of the imitation gap on learning driving strategies. By introducing partial observability constraints, the dataset helps evaluate and improve methods for learning autonomous driving strategies from human driving demonstrations.

- 1IGDrivSim: A Benchmark for the Imitation Gap in Autonomous Driving索邦大学 · 2024年



