CapsuLSTM-BOA: A Hybrid Deep Learning and Bio-Inspired Routing Model for Mobile Ad-Hoc Network

Abstract

Establishing consistent, adaptable connectivity in infrastructure-free situations is a significant challenge for decentralized mobile communication. This research article presents a unique hybrid routing system that blends bio-inspired optimization and Deep Learning (DL) for Android-based Mobile Ad-Hoc Networks (MANETs). Proposed framework combines Long Short-Term Memory (LSTM) network for describing temporal fluctuations in node mobility, link reliability, and traffic patterns, with a Capsule Network to extract high-level spatial correlations in network architecture. The system uses the Bonobo Optimization Algorithm (BOA), which is modelled after the social intelligence and cooperative behavior of bonobos, to choose the best routing paths dynamically. By striking a balance between exploration and exploitation, BOA assesses routing choices based on important variables such as transmission delay, node movement, energy levels, and link quality. Even in the face of quickly shifting topologies, highly adaptive, real-time peer-to-peer communication is made possible by the combination of CapsuLSTM and BOA. According to simulation results, the proposed CapsuLSTM-BOA framework offers a potent solution for file sharing, mobile collaboration, and resilient communication in decentralized networks, outperforming conventional MANET routing protocols in terms of end-to-end delay, throughput, Packet Delivery Ration (PDR), packet loss ratio, energy consumption, end-to-end delay, routing control overhead and network lifetime.