Radial Bias Function Neural Network and Packet Chaining Reservation Protocol for Congestion Prediction and Avoidance in MANET under TCP Traffic Patterns

The effectiveness of Transmission Control Protocol (TCP) connections is significantly impacted by congestion control (CC). Congestion occurs when the network cannot manage the level of traffic that arises when a large number of packets arrive at once. If congestion is anticipated, packet loss can be prevented by taking appropriate measures to lower the rate of packet production at the source. Many existing protocols are developed to control congestion on the other hand, the packet delivery ratio will not sufficiently be high and it lengthens the delay, which affects the performance of TCP. In order to improve TCP performance in MANET, the Packet Chaining Reservation Protocol (PCRP) using radial basis functional neural network (RBFNN) was developed. The initial message is transmitted by the transmitter, then waits for a reply. If the acknowledgement is received, the congestion prediction algorithm is executed to determine whether there is congestion or not on the data transmission line. Based on a few characteristics, RBFNN is used to estimate congestion in the transport layer. The best size of the cwnd is chosen using the osprey optimization algorithm (OOA), which is used for efficient data transfer. If the cwnd value above the threshold value signifies congestion avoidance phases are carried out in accordance with PCRP. This process is repeated until every message has been transmitted. The simulation analysis shows the proposed protocol has 89 $$\%$$ packet delivery ratio, 2.5Mbps of throughput and 15 sec delay. Thus, the proposed approach is the better choice for enhancing the performance of TCP under MANET.