Dataset Title: Payload Configurations
Dataset Description: Geographical traffic demand and corresponding payload configuration in terms of power and bandwidth for each beam.
The beam pattern was kindly provided by the European Space Agency (ESA). ESA has given permission to SnT to share this dataset.
This dataset has been used in this work (please cite this reference in your work if you make use of this dataset):
- F. Ortiz, N. Skatchkovsky, E. Lagunas, W. Martins, G. Eappen, S. Daoud, O. Simeone, B. Rajendran, S. Chatzinotas, “Energy-Efficient On-Board Radio Resource Management for Satellite Communications via Neuromorphic Computing”, under revision to IEEE Trans. On Machine Learning in Communications and Networking, Nov-2023.
Channel per beam and users positions
Same beam pattern used here.
Dataset Description:
This dataset, titled “Payload Configurations,” is an extensive collection of data designed for SatCom machine learning research. It encompasses geographical traffic demand and corresponding payload configurations, detailed in terms of power and bandwidth for each service beam.
Key Features:
1. Geographical Coverage and Beam Pattern: The dataset includes eight beams covering specific geographical coordinates (latitude and longitude). This pattern is provided by the European Space Agency (ESA), with authorization to share the data. The dataset includes data from eight beams aimed at specific geographical positions:
- Beam 1: Latitude 39.3, Longitude -5.3
- Beam 2: Latitude 42, Longitude 0
- Beam 3: Latitude 44.7, Longitude 5.3
- Beam 4: Latitude 47.4, Longitude 10.6
- Beam 5: Latitude 51, Longitude -0.5
- Beam 6: Latitude 53.7, Longitude 6
- Beam 7: Latitude 56.4, Longitude 12.3
- Beam 8: Latitude 39.5, Longitude 14.4
2. Pre-processing and Simulation: A traffic simulator pre-processes the data, eliminating redundant information, resolving gaps, and categorizing User Terminals (UTs) according to their geographic positions. This process involves the utilization of flight data and maritime traffic data for modeling daily and hourly demands.
3. Training Data Structure: The training data comprises demand requests formatted in matrix form, representing the traffic demand across a grid of latitude and longitude points within the service area. Additionally, labels for these data points are generated through exhaustive search methods to determine optimal power and bandwidth allocations, thereby aiding in efficient payload resource management.
4. Payload Configuration Representation: Each payload configuration is a combination of feasible power and bandwidth pairs for each beam. Due to constraints in total power and bandwidth, not all configurations are viable, and the dataset reflects this.
5. Data Generation and Computational Resources: The data generation process, conducted using the University of Luxembourg’s HPC facilities, is intensive, with significant time investment per sample to ensure optimal resource allocation.
6. Data Structure and Storage: The dataset is stored in three primary variables:
- `data.mat`: Contains geographic demand with dimensions of 381x651x30000 (longitude points x latitude points x number of samples).
- `Payload_configuration`: A vector of dimensions 1×30000, representing the payload configurations per sample.
- `Capacity_beam.mat`: Details the capacity assigned per beam in a dimension of 8×30000 (number of beams x number of samples).