DL-TCP: Deep Learning-Based Transmission Control Protocol for Disaster 5G mmWave Networks

Abstract

The 5G mobile communication system is attracting attention as one of the most suitable communication models for broadcasting and managing disaster situations, owing to its large capacity and low latency. High-quality videos taken by a drone, which is an embedded IoT device for shooting in a disaster environment, play an important role in managing the disaster. However, the 5G mmWave frequency band is susceptible to obstacles and has beam misalignment problems, severing the connection and greatly affecting the degradation of TCP performance. This problem becomes even more serious in high-mobility drones and disaster sites with many obstacles. To solve this problem, we propose a deep-learning-based TCP (DL-TCP) for a disaster 5G mmWave network. DL-TCP learns the node's mobility information and signal strength, and adjusts the TCP congestion window by predicting when the network is disconnected and reconnected. As a result of the experiment, DL-TCP provides better network stability and higher network throughput than the existing TCP NewReno, TCP Cubic, and TCP BBR.

Keywords

• Deep-learning
• mmWave
• TCP
• 5G
• supervised-learning