Adaptive Routing Strategies for Optimisation in MANETs Through Integration of Expanding Ring Search and Random Early Detection Using Machine Learning

This work presents a novel machine learning (ML)-driven framework to optimise reactive routing protocols (RRPs) in mobile ad hoc networks (MANETs), tackling congestion control through intelligent, real-time protocol selection. Building on our prior Adaptive Expanding Ring Search (AERS) method enhanced with random early detection (RED), the study introduces an ML classification system that dynamically identifies the most efficient RRP based on network conditions. High-accuracy classifiers, AdaBoost (95%), K-Nearest Neighbours (93%), and Decision Trees (92%), enable data-driven decision-making, systematically evaluating protocols across diverse topologies to maximise performance. The framework ensures context-aware routing, significantly improving Quality of Service (QoS) through enhanced packet delivery, reduced latency, and robust congestion mitigation. Rigorous NS-3 simulations validate the approach, demonstrating measurable gains over conventional methods. By integrating predictive analytics into routing strategy, this research advances the design and deployment of RRPs, bridging algorithmic innovation with practical implementation. The results offer high-impact insights for both academic research and real-world MANET applications, establishing a new paradigm for adaptive, efficient routing in dynamic wireless environments.