/deep-learning-based-radioMap

This is the repository for ICC 'UAV-aided Joint Radio Map and 3D Environment Reconstruction using Deep Learning Approaches' by Junting Chen and Pengxi Zeng.

Primary LanguagePythonGNU General Public License v2.0GPL-2.0

modules

Deep learning models.

  • algorithmSim.py: The main frame of the whole net, combining all the solitary modules. Multichannel Gain and the Obstruction Network are implemented directly in the main frame class Algorithm.
  • Convolutional_RE_Net.py: Convolutional RE-Net is realized.
  • weightGen.py: The two nonlinear models in the Obstruction Network and the RM-Net are realized here.
  • Linear_RE_Net.py: Linear RE-Net is realized.

utils

Some useful functions and settings.

  • Config.py: Basic settings.
  • dist.py: Calculate the log-distance.
  • line_calc.py: Calculate the line equation.
  • load_dict.py: Load the trained part into the combined framework.
  • Map.py: Transform the location pair to a position map.

dataset

Generate the dataset.

  • locPair.py: Obtain the location pair and the RSS measurements.

Running Files

  • training.py: Train our framework.
  • weight_pretrain: Pre-train the nonlinear models in the Obstruction Network and the RM-Net.
  • performance_evaluation.py: Evaluate the performance of our framework with whole datasets.

Tactic

  • Set the basic settings, like the dataset filename in training.py, all the parameters in the Config.py.
  • Initialize the parameters of Multichannel Gain in algorithmSim.py.
  • Run weight_pretrain.py.
  • Run training.py.
  • Run performance_evaluation.py to see the performance.