Airborne Network Communication

This dataset contains processed simulation results from an Airborne Network. The simulations involved networks of 25 to 50 aerial vehicles with varying network parameters. In each simulation, 50 to 100 data flows with diverse source-destination pairs were scheduled for transmission. These flows consisted of 5 to 50 packets each, with packet sizes ranging from 300 to 1200 bytes, and were transmitted at 0.5-second intervals.
During network initialization, the distance between nodes followed a normal distribution with a specified mean (e.g., 160 km) and standard deviation, bounded by a minimum of 1 km and a maximum of 300 km. Links were established based on hop-to-hop distance, with a maximum reception range of 250 km. Following the simulations, a PyTorch dataset was created to facilitate the development of a model for estimating experienced delay based on the current network state and determined route. The models under consideration include Graph Neural Networks (GNNs), Feed-Forward Neural Networks (FFNs), Random Forest, Support Vector Machines, and Gradient Boosting Regressors (GBRs). The dataset comprises input features representing the network state and the corresponding output, which is the experienced delay.
For GNN-based algorithms, the input is PyTorch Geometric data consisting of node hidden states and adjacency indices. In this representation, GNN nodes correspond to network links and data flows. Four distinct datasets were generated, each with different hidden state dimensions and included features. The hidden state dimension exceeds the number of included features, with the remaining dimensions padded with zeros. These datasets are:
dataset96_1: 96-dimensional hidden states, including 1 feature: link data rate (for links) and packet size (for flows);
dataset96_3: 96-dimensional hidden states, including 3 features: link data rate, propagation delay, and link velocity (for links); and packet size and number of scheduled packets (for flows);
dataset64_1: 64-dimensional hidden states, including 1 feature: link data rate (for links) and packet size (for flows);
dataset64_3: 64-dimensional hidden states, including 3 features: link data rate, propagation delay, and link velocity (for links); and packet size and number of scheduled packets (for flows).
For machine learning-based algorithms, the input is a PyTorch tensor containing: the average data rate of links in the path, the minimum data rate in the path, the total propagation delay, the number of hops in the path, the packet size of the flow, the number of packets in the flow, the current maximum distance between two hops, and the link velocity of the link with the maximum distance.