DataAI-6G: A System Parameters Configurable Channel Dataset for AI-6G Research

Zibing Shen,Jianhua Zhang,Li Yu,Yuxiang Zhang,Zhen Zhang,Xidong Hu

2023-06-03

Abstract:With the acceleration of the commercialization of fifth generation (5G) mobile communication technology and the research for 6G communication systems, the communication system has the characteristics of high frequency, multi-band, high speed movement of users and large antenna array. These bring many difficulties to obtain accurate channel state information (CSI), which makes the performance of traditional communication methods be greatly restricted. Therefore, there has been a lot of interest in using artificial intelligence (AI) instead of traditional methods to improve performance. A common and accurate dataset is essential for the research of AI communication. However, the common datasets nowadays still lack some important features, such as mobile features, spatial non-stationary features etc. To address these issues, we give a dataset for future 6G communication. In this dataset, we address these issues with specific simulation methods and accompanying code processing.

Signal Processing

What problem does this paper attempt to address?

The main objective of this paper is to address the issue of accurately obtaining Channel State Information (CSI) in future 6G communication systems. Specifically, with the accelerated commercialization of 5G technology and the research progress of 6G communication systems, communication systems exhibit characteristics such as high frequency, multi-band, high-speed user mobility, and large-scale antenna arrays. These characteristics make it difficult for traditional methods to obtain accurate channel state information, thereby limiting the performance of traditional communication methods.

To solve this problem, researchers propose using Artificial Intelligence (AI) technology to replace or supplement traditional modeling methods. The paper introduces a dataset named DataAI-6G, which aims to support machine learning research in wireless channel transmission and modeling. This dataset has the following features:

1. **Mobility Characteristics**: The dataset considers the mobility function of users and introduces the Doppler effect to simulate channel state information under mobile conditions.
2. **Spatial Non-Stationary Characteristics**: The dataset considers near-field spatial non-stationary characteristics during simulation, which is an important feature in 6G communication.
3. **Flexible Configuration**: The dataset allows researchers to freely configure user paths and speeds, enabling the generation of any number of user points.

With the above features, the DataAI-6G dataset can better adapt to existing algorithms and provide more realistic channel characteristics. Additionally, the paper uses a beam prediction algorithm to verify the effectiveness of this dataset. Experimental results show that models trained using the DataAI-6G dataset have higher prediction accuracy at different speeds compared to models using other datasets.