nuPlan-v1.0 dataset

nuPlan是世界上第一个自动驾驶的大型规划基准。尽管越来越多的基于ML的运动计划人员，但缺乏已建立的数据集，仿真框架和指标限制了该领域的进展。用于自动驾驶车辆运动预测的现有基准 (Argoverse，Lyft，Waymo) 集中在其他代理的短期运动预测上，而不是自我车辆的长期计划。这导致以前的工作使用具有L2-based指标的开环评估，这些指标不适合公平评估长期计划。我们的基准测试通过提供培训框架来开发基于机器学习的计划者，轻量级闭环模拟器，特定于运动计划的指标以及可视化结果的交互式工具，克服了这些限制。

nuPlan is the world's first large-scale planning benchmark for autonomous driving. Despite the growing number of ML-based motion planners, the lack of established datasets, simulation frameworks, and metrics has hindered progress in this field. Existing benchmarks for autonomous driving motion prediction (Argoverse, Lyft, Waymo) focus on short-term motion prediction of other agents rather than long-term planning of the ego vehicle. As a result, prior work has relied on open-loop evaluation with L2-based metrics, which are not suitable for fair evaluation of long-term planning. Our benchmark overcomes these limitations by providing a training framework for developing ML-based planners, a lightweight closed-loop simulator, motion planning-specific metrics, and interactive tools for visualizing results.

搜集汇总

背景与挑战

背景概述

nuPlan-v1.0数据集是世界上第一个自动驾驶大型规划基准，专注于长期运动规划而非短期预测，以克服现有基准的局限性。它提供了完整的培训框架、轻量级闭环模拟器、专用指标和可视化工具，旨在推动基于机器学习的规划器发展。该数据集由Motional于2021年6月发布，大小为1371.1GB，采用Apache 2.0许可证。