MIMO Systems Channel Estimation with GNNs
Budget: ₹1,500 – ₹12,500 INR
The objective of this project is to develop a novel Graph Neural Network (GNN) architecture to predict the real and imaginary components of Channel State Information (CSI) in MIMO systems. The goal is to leverage the spatial and temporal relationships between subcarriers and antenna pairs for enhanced channel estimation accuracy, while also reducing computational complexity. We aim to outperform traditional models like RNNs or Transformers.
Key details:
- The dataset is structured as [1000, 624, 4, 4, 2], which includes 1000 samples, 624 subcarriers, and 4x4 transmit-receive antennas with corresponding real and imaginary components.
- To simplify the problem, the dataset is downsampled to [1000, 52, 4, 4, 2].
- Each graph node corresponds to a subcarrier, with features encoding the CSI for transmit-receive pairs. Edges in the graph represent spatial or channel dependencies.
The ideal candidate for this project will have:
- Extensive experience with Graph Neural Networks and their application in channel estimation.
- Proficiency in data preprocessing techniques including normalization and downsampling.
- Ability to design and implement a GNN model that outputs combined real and imaginary components.
Please note that both enhanced channel estimation and reduced computational complexity are equally important objectives for this project.
Key details:
- The dataset is structured as [1000, 624, 4, 4, 2], which includes 1000 samples, 624 subcarriers, and 4x4 transmit-receive antennas with corresponding real and imaginary components.
- To simplify the problem, the dataset is downsampled to [1000, 52, 4, 4, 2].
- Each graph node corresponds to a subcarrier, with features encoding the CSI for transmit-receive pairs. Edges in the graph represent spatial or channel dependencies.
The ideal candidate for this project will have:
- Extensive experience with Graph Neural Networks and their application in channel estimation.
- Proficiency in data preprocessing techniques including normalization and downsampling.
- Ability to design and implement a GNN model that outputs combined real and imaginary components.
Please note that both enhanced channel estimation and reduced computational complexity are equally important objectives for this project.