TCP-CNNLSTM: Congestion Control Scheme for MANET using AI Technologies

The primary focus of the research is on how artificial intelligence techniques may be applied to MANETs to improve communication and safe routing between mobile dynamic nodes. In this paper, a proposed machine learning-based congestion control method, the TCP-CNNLSTM model. This model works in the classification and decision-making process. It collects network traffic traces, selects a few factors, and then picks the CNN and LSTM models to train the model. TCP-CNNLSTM is a model based on deep learning that learns the most effective congestion management method. The Kanerva coding function approximation is utilized to decrease the complexity of the system as well as the size of the state vector and action value. Experimental results have provided high throughput and minimized RTT, packet loss, and ACK Interval. It has also been compared with state-of-the-art RL-based the NewReno method and congestion control methods. In this comparison, the proposed TCP-CNN-LSTM model outperforms state-of-the-art methods.