AirPlayground: Benchmarking Aerial Navigation across Diverse Reconstructed Environments

📄 arXiv: 2610.29699v1 📥 PDF

作者: Rangya Zhang, Yujing Shang, Zhuoyuan Yu, Mingsheng Li, Qingsong Yan, Chao Yan, Yuhang Zhang, Mir Feroskhan

分类: cs.RO

发布日期: 2026-10-21


💡 一句话要点

提出AirPlayground以解决无人机导航评估与训练问题

🎯 匹配领域: 支柱三:空间感知与语义 (Perception & Semantics)

关键词: 无人机导航 3D环境构建 自动化飞行收集 策略训练 闭环评估

📋 核心要点

  1. 现有的无人机导航方法在环境覆盖和飞行演示的密集采样方面存在不足,限制了策略学习的有效性。
  2. 本研究提出AirPlayground基准,通过构建多样化的3D环境和自动化飞行收集管道,提供丰富的训练数据和评估标准。
  3. 实验结果表明,扩大训练语料库显著提高了在已见和未见环境中的成功率,验证了方法的有效性。

📝 摘要(中文)

大规模的空中资源为学习不同环境下的导航提供了基础。然而,现有资源在提供密集采样的飞行演示和广泛环境覆盖方面仍然有限。为了解决这些问题,我们提出了AirPlayground,这是一个统一的基准,用于在多样化环境中训练和评估无人机导航策略。我们构建了94个逼真的3D环境,开发了自动化专家飞行收集的场景自适应管道,生成了94K演示和4650万同步观察-状态-动作样本。我们还提供了一个可扩展的API,支持在共享语料库上训练不同的策略,并在相应环境中进行统一评估。通过这一框架,AirPlayground支持分析离线预测误差与闭环结果之间的关系。

🔬 方法详解

问题定义:本论文旨在解决现有无人机导航资源在环境覆盖和飞行演示密度不足的问题,限制了策略学习和评估的有效性。

核心思路:通过构建多样化的3D环境和开发自动化的飞行收集管道,AirPlayground为无人机导航策略的训练和评估提供了统一的基准和丰富的数据支持。

技术框架:整体架构包括三个主要模块:1) 94个逼真的3D环境构建,涵盖室内、城市和植被场景;2) 场景自适应的自动飞行收集管道,生成94K飞行演示;3) 可扩展的API,支持不同策略的训练和评估。

关键创新:AirPlayground的核心创新在于其统一的基准和丰富的环境覆盖,解决了现有方法在数据稀缺和评估标准不统一方面的不足。

关键设计:在技术细节上,采用了高频率(50Hz)的数据采集,确保了样本的同步性和准确性,同时设计了适应不同策略的训练和评估协议。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,通过将ACT的训练语料库从5个场景扩展到70个场景,已见和未见环境中的成功率分别提高至41.2%和29.4%,相较于原先的30.6%和16.8%有显著提升,验证了方法的有效性和可扩展性。

🎯 应用场景

AirPlayground的研究成果在无人机导航、自动驾驶和机器人领域具有广泛的应用潜力。通过提供标准化的训练和评估框架,研究者和工程师可以更有效地开发和优化导航策略,推动相关技术的进步与应用。

📄 摘要(原文)

Large-scale aerial resources provide a foundation for learning navigation across diverse environments. However, existing resources remain limited in providing (1) densely sampled flight demonstrations for learning different policy formulations and (2) broad environment coverage for standardized closed-loop evaluation. To address these limitations, we introduce AirPlayground, a unified benchmark for training and evaluating aerial navigation policies across diverse environments. We first construct 94 photorealistic 3D Gaussian Splatting (3DGS) environments spanning indoor, urban outdoor, and vegetation settings, with aligned collision geometry. Building on these environments, we develop a scene-adaptive pipeline for automated expert flight collection, yielding 94K demonstrations and 46.5M synchronized observation-state-action samples at 50 Hz. We further provide an extensible API that supports training different policy formulations on this shared corpus and evaluating them under unified protocols in the corresponding environments. We benchmark seven baselines spanning four policy paradigms across three navigation settings in seen and unseen environments, and further examine scaling behavior with ACT. Scaling ACT's training corpus from 5 to 70 scenes raises median success rates on seen and unseen environments from 30.6% and 16.8% to 41.2% and 29.4%, respectively. Through this shared training and evaluation framework, AirPlayground supports analysis of how offline prediction errors relate to closed-loop outcomes. Our analyses highlight the roles of prediction errors, execution protocols, and scene interactions in closed-loop performance.