无线AI研究数据集

1. 数据集简介 无线AI研究数据集基于真实地图场景构建，具有场景丰富、参数灵活可配、简单易用等特点。本数据集能够支持多种无线AI任务，包括定位和环境重建等传感任务、RIS和波束等MIMO任务、CSI反馈和信道估计等物理层任务。 场景丰富：从全球40多个大城市的真实地图中随机挑选了10000个场景，建筑物布局更加符合实际场景。 参数灵活可配：提供了射线追踪产生的射线参数，支持用户自定义通信参数，生成无线AI任务所需的数据。 简单易用：提供了数据生成代码、数据预处理代码和示例任务训练代码，方便用户快速上手使用。 2. 数据集结构 针对不同场景和设备撒点密度，数据集包括两部分： Part 1：丰富场景、稀疏部署（10000张地图场景，每个场景5个基站位置，30个UE位置） Part 2：特定场景、密集部署（100张地图场景，每个场景1个基站位置，10000个UE位置） 图1：建筑物和BS/UE俯视图 图2：建筑物鸟瞰图 详细的数据集信息请参考附件“无线AI研究数据集 README.docx”。

1. Dataset Introduction
This wireless AI research dataset is constructed based on real-world map scenarios, featuring rich scenarios, flexible and configurable parameters, and ease of use. It supports a variety of wireless AI tasks, including sensing tasks such as positioning and environment reconstruction, MIMO tasks such as RIS and beamforming, and physical layer tasks such as CSI feedback and channel estimation.

Rich Scenarios: 10,000 scenarios are randomly selected from real maps of over 40 major cities worldwide, with building layouts that better match actual real-world scenarios.
Flexible and Configurable Parameters: Ray parameters generated via ray tracing are provided, allowing users to customize communication parameters to generate data required for wireless AI tasks.
Ease of Use: Data generation code, data preprocessing code, and example task training code are provided to facilitate users' quick adoption and usage.

2. Dataset Structure
Based on different scenarios and device deployment densities, the dataset consists of two parts:
Part 1: Rich Scenarios, Sparse Deployment (10,000 map scenarios, with 5 base station (BS) locations and 30 User Equipment (UE) locations per scenario)
Part 2: Specific Scenarios, Dense Deployment (100 map scenarios, with 1 base station location and 10,000 UE locations per scenario)

Figure 1: Top-down view of buildings and BS/UE locations
Figure 2: Bird's-eye view of buildings

For detailed dataset information, please refer to the attached document "Wireless AI Research Dataset README.docx".

搜集汇总

背景与挑战

背景概述

无线AI研究数据集是一个基于真实地图场景构建的综合性数据集，包含全球40多个城市的10000个场景，支持多种无线AI任务。数据集分为稀疏部署和密集部署两部分，提供灵活的参数配置和易用的代码支持，适合定位、环境重建、MIMO等研究需求。