Skip to content

Code for a class project on predicting future blockage events for a 6G link based on past received power levels

Notifications You must be signed in to change notification settings

5-Jeremy/ECEN683-Blockage-Prediction

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Blockage Prediction for Wireless Communication Using mmWave Power Signatures

This project is an attempt to solve the task of Wireless Signature-Based Blockage Prediction proposed by the creators of the DeepSense 6G dataset. The task considers a wireless transmitter sending a constant stream of symbols to a receiver, which measures the received power for a codebook of 64 beams. Over time, obstacles briefly enter the space between the transmitter and receiver, resulting in a blockage which prevents the symbols from being received. A deep learning model is provided a sequence of mmWave samples collected in this way, and must output a prediction for whether a blockage will occur over a certain number of future time steps.

The use of a convolutional neural network (CNN) or a recurrent neural network (RNN) individually to solve this task was explored in a paper by the creators of the DeepSense 6G dataset. The code presented here combines a convolutional layer with an RNN and applies it to the same task.

For a complete presentation of the project including the hyperparameter settings I used and the results I obtained, see this video.

Before you can use the file train_and_test.ipynb to replicate my results or experiment with the settings, you need to download scenarios 17 through 21 from the DeepSense 6G dataset here. The files need to be extracted and placed within the same directory as train_and_test.ipynb and prep_data.py. I recommend deleting all of the images in camera_data folder because they take up a lot of space and are not used by this model.

About

Code for a class project on predicting future blockage events for a 6G link based on past received power levels

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published