A Dynamic and Adaptive Framework for Efficient Configuration and Management in Mobile Ad Hoc Networks

Abstract

Mobile Ad Hoc Networks (MANETs) often face challenges such as frequent topology changes, limited resources, and security vulnerabilities, which affect their stability and performance. This study introduces DynaMANET, a framework designed to address these challenges by integrating routing, resource management, and address configuration. The framework employs a multi-layered approach: the Data Collection Layer gathers network metrics, the Learning and Decision-Making Layer uses Deep Q-Networks to adjust configurations, and the Context Awareness Layer applies Graph Neural Networks to detect network changes. The Feedback and Evaluation Layer utilizes Bayesian optimization to refine decisions, while the Execution and Control Layer ensures seamless implementation of configurations. Simulations were performed in a 100-node network environment modeled using the NS-3 simulator. Node mobility and varying traffic loads were considered to evaluate throughput, latency, packet delivery ratio, energy consumption, and intrusion detection rates. The framework demonstrated better throughput, reduced delays, and higher packet delivery ratios compared to E-OLSR and MARL models. Detection rates for security threats such as black hole and DDoS attacks were slightly higher. Scalability tests showed minimal throughput drop as the network size increased. In failure scenarios, the framework maintained stable communication and data delivery. These results suggest that DynaMANET addresses interconnected challenges in MANETs by combining adaptive decision-making with efficient resource management. The methods proposed provide a structured approach for improving MANET reliability under dynamic conditions.