Data underlying the publication: MAVRL: Learn to Fly in Cluttered Environments with Varying Speed

Datasets for paper of 'MAVRL: Learn to Fly in Cluttered Environments with Varying Speed'. This research is about how to fly in a cluttered environment based on reinforcement learning. There are 5 .zip files in this dataset: vae.zip is the training result of Variational AutoEncoder; lstm_dataset.zip is the collected data of many rollouts by an initial policy; RecurrentPPO_1.zip contains a initial policy and retrained policy; LSTM_110_9_0.zip is the training result of LSTM; AvoidBench.zip is the unity standalone file.本数据集对应论文《MAVRL：面向复杂变速环境的飞行学习》（英文原名：MAVRL: Learn to Fly in Cluttered Environments with Varying Speed）。该项研究聚焦于基于强化学习的复杂环境自主飞行任务。本数据集共包含5个压缩文件：vae.zip为变分自编码器（Variational AutoEncoder）的训练结果；lstm_dataset.zip为初始策略采集的多回合交互轨迹数据；RecurrentPPO_1.zip包含初始策略与重训练策略；LSTM_110_9_0.zip为长短期记忆网络（Long Short-Term Memory，LSTM）的训练结果；AvoidBench.zip为Unity独立运行文件。

搜集汇总

背景与挑战

背景概述

该数据集支持一篇关于无人机在杂乱环境中基于强化学习飞行的研究论文，包含5个.zip文件，涵盖变分自编码器、LSTM训练结果、策略数据及Unity模拟文件，总大小约4GB。数据集聚焦于人工智能和航空航天工程领域，关键词包括无人机、强化学习和障碍物回避，使用CC BY 4.0许可证开放共享。