[Paper Review] DeepMIMO: A Generic Deep Learning Dataset for Millimeter Wave and Massive MIMO Applications

Ahmed Alkhateeb|arXiv (Cornell University)|Feb 18, 2019

Millimeter-Wave Propagation and ModelingEngineering259 citations

TL;DR

DeepMIMO provides a ray-tracing–based, parametric dataset for mmWave/massive MIMO ML research, enabling reproducible benchmarking; an outdoor example with 18 BS and over a million users demonstrates beam prediction use.

ABSTRACT

Machine learning tools are finding interesting applications in millimeter wave (mmWave) and massive MIMO systems. This is mainly thanks to their powerful capabilities in learning unknown models and tackling hard optimization problems. To advance the machine learning research in mmWave/massive MIMO, however, there is a need for a common dataset. This dataset can be used to evaluate the developed algorithms, reproduce the results, set benchmarks, and compare the different solutions. In this work, we introduce the DeepMIMO dataset, which is a generic dataset for mmWave/massive MIMO channels. The DeepMIMO dataset generation framework has two important features. First, the DeepMIMO channels are constructed based on accurate ray-tracing data obtained from Remcom Wireless InSite. The DeepMIMO channels, therefore, capture the dependence on the environment geometry/materials and transmitter/receiver locations, which is essential for several machine learning applications. Second, the DeepMIMO dataset is generic/parameterized as the researcher can adjust a set of system and channel parameters to tailor the generated DeepMIMO dataset for the target machine learning application. The DeepMIMO dataset can then be completely defined by the (i) the adopted ray-tracing scenario and (ii) the set of parameters, which enables the accurate definition and reproduction of the dataset. In this paper, an example DeepMIMO dataset is described based on an outdoor ray-tracing scenario of 18 base stations and more than one million users. The paper also shows how this dataset can be used in an example deep learning application of mmWave beam prediction.

Motivation & Objective

• Motivate the need for a large, environment-aware dataset for mmWave/massive MIMO ML research.
• Propose a generic, parameterized dataset generation framework that captures environment geometry and transmitter/receiver locations.
• Ensure reproducibility by making the dataset completely defined by the ray-tracing scenario and parameter set.
• Demonstrate the dataset through an example application in mmWave beam prediction.
• Provide a practical workflow and code to generate and use the dataset for ML tasks.

Proposed method

• Construct channels from accurate ray-tracing outputs (Wireless InSite) to capture environment dependence.
• Define a parametric dataset framework with a scenario R and parameter set S to tailor datasets.
• Compute OFDM-channel vectors h_k^{b,u} across selected subcarriers using a sum of L paths with array responses a(·) (Eq. 1–5).
• Include user location and other features to enable ML inputs beyond channels.
• Output a MATLAB MAT-file DeepMIMO_dataset.mat with structured access to channels and user locations.
• Provide a detailed workflow to generate datasets for chosen scenarios and parameter settings.

Experimental results

Research questions

• RQ1 How can a large, environment-aware, parameterizable dataset be constructed for mmWave/MIMO ML research?
• RQ2 How can ray-tracing outputs be transformed into ML-ready channel inputs and beamforming targets?
• RQ3 What is the impact of dataset parameters (antennas, bandwidth, OFDM, number of paths) on learning tasks like beam prediction?
• RQ4 Can DeepMIMO enable reproducible benchmarking across different ML approaches in mmWave/MIMO?

Key findings

• The DeepMIMO dataset is driven by accurate ray-tracing data, capturing environment geometry and transmitter/receiver locations.
• The framework is generic and parametric, allowing researchers to tailor datasets to their ML tasks by adjusting S and selecting scenarios R.
• An example DeepMIMO dataset (O1) includes 18 base stations and over one million users, facilitating large-scale ML experiments.
• The dataset supports constructing inputs/outputs for supervised learning applications such as mmWave beam prediction.
• The paper demonstrates a beam prediction application that uses Omni received sequences at multiple BSs to predict beamforming vectors.
• The dataset and accompanying code enable reproducibility and straightforward comparison across ML methods.