The 5G mobile telecommunication network is becoming known as one of the finest communication networks for transmitting and controlling emergencies due to its high bandwidth and short latency. The high-quality videos taken by a drone, an incorporated Internet of Things (IoT) gadget for recording in a catastrophe situation, are very helpful in controlling the crisis. The 5G mm-Wave frequency spectrum is susceptible to intrusion and has beam realignment concerns, which can severely reduce Transmission Control Protocol (TCP) effectiveness and destroy connections. High-speed devices and disaster zones with multiple barriers make this problem significantly worse. This research offers a Deep-Learning-oriented Congestion Control Approach (DLCCA) for a catastrophic 5G mm-Wave system to solve this problem. By analyzing the node's transmitted data, DLCCA predicts when the network will be disconnected and reconnected, adjusting the TCP congestion window accordingly. To accomplish this, the proposed approach estimates the bottleneck link's queue length using the average Round Trip Time (RTT) and its data collected across the connection.
Consequently, the proposed approach can use this buffer size to examine the congestion state and differentiate it from the randomized loss situation. This would stop the window length from getting smaller, increasing the amount of data transferred and speeding up the recommended method. Additionally, DLCCA frees up bottleneck bandwidth. The research provides simulated tests for TCP DLCCA compared to Newreno, Cubic, Compound, and Westwood while sustaining a two-way connection under heavy load and a wide range of randomized loss rates using the networking simulation NS-2. Experimental results show that DLCCA performs better than other TCP variants and significantly boosts throughput.