Line-of-sight Blockage Prediction using CNN-LSTM pipelines in 6G communication systems.

What is it about?

The future 6G communications systems will require extremely fast data transfer at ultra-low latency with high reliability. Hence, predicting the future line-of-sight blockages accurately becomes an imperative task to avoid link disruptions in such systems. We have developed CNN-LSTM pipelines to capture the spatial-temporal features and predict future line-of-sight blockages efficiently using sequential image frames captured by a RGB camera.

Featured Image

Why is it important?

The proposed deep learning architecture, using CNN for feature extraction and LSTM for future blockage prediction, demonstrated improved performance compared to the original model that employed Gated Recurrent Units for future blockage prediction and YOLOv3 for object detection. The evaluation metrics reveal significant improvements, with an average increase of 10.63% in Top-1 accuracy for the ResNet-LSTM model and 12.60% for the ResNeXt-LSTM model compared to the baseline. Furthermore, the F1 score has also improved by an average of 7.58% and 8.13% for the ResNet-LSTM and ResNeXt-LSTM models. The key performance indicators improvement demonstrates that the proposed model results in an efficient, and reliable wireless communication system.