M-DODGE: Mutation Based Dragon Fly Optimized Deep Learning Model for Congestion Control in MANET
R. Dinesh, A. Ahilan, Narayanaperumal Muthukumaran, S. Gladson · IETE Journal of Research · 2025
Mobile Ad hoc Network (MANET) is made up of mobile devices that form an infrastructure-less network. Because of the fast proliferation of mobile devices, a huge number of messages are sent during information exchange in congested locations. It can cause congestion that results in network delay, low throughput, data packet loss and poor energy efficiency. To overcome these challenges this paper proposes a novel Mutation based Dragonfly Optimized Deep learning for conGestion Elimination (M-DODGE) approach that detects the congested node effectively. The MANET data such as Transmission Control Protocol (TCP) and User Datagram Protocol (UDP) undergoes preprocessing to remove the missing values and normalize the data. After preprocessing the M-DODGE approach employs Mutation-fused Dragonfly Optimization (MDO) for effective feature selection and enhances the accuracy. The suggested approach utilizes a hybrid Spiking Convolutional Neural Network based Bidirectional Gated Recurrent Unit (SCNN-BiGRU) model for accurate congestion node detection. The efficacy of M-DODGE was compared with BiLSTM, BiGRU, and LSTM utilizing the Ns-2 simulator. The efficacy of the suggested M-DODGE framework is evaluated utilizing several metrics namely Delay (DE), Packet Delivery Ratio (PDR), Energy Consumption (EC), Throughput (TP), Network Lifetime (NL), and Detection Rate (DR). The SCNN-BiGRU model enhances the overall accuracy by 11.28%, 9.76%, 12.5% and 3.84% over the existing BiLSTM, BiGRU, CNN, and LSTM techniques respectively.