Design of an Iterative Model Integrating Bacterial Foraging Optimizer and Q-Learning for Enhanced Congestion Management in Wireless Networks

Main Article Content

Chandrashekhar K.Beral , Manish N. Tibdewal

Abstract

Abstract: In the realm of wireless networks, congestion management poses a critical challenge due to the dynamic and dense nature of modern network traffic. Existing methods, which predominantly utilize traditional routing and heuristic-based strategies, often fall short in adaptability and efficiency, leading to degraded service quality and increased latency. This work introduces a novel, integrated approach combining Bacterial Foraging Optimizer (BFO), ensemble classification via Multi-Layer Perceptron (MLP) and Logistic Regression (LR), and Q-Learning, which collectively address these limitations through enhanced path optimization and congestion detection capabilities. The proposed model leverages the Bacterial Foraging Optimizer for path finding, utilizing its bio-inspired mechanisms to emulate the natural foraging behaviors of bacteria, thereby efficiently navigating the solution space to identify less congested routes. This is complemented by the use of an ensemble classification system combining MLP and LR. MLP excels in capturing complex nonlinear relationships within data, whereas LR provides insightful probabilistic outputs, enhancing the overall accuracy in predicting congested nodes and paths. The integration of these classifiers helps in reliably identifying congestion in varying network scenarios. Furthermore, Q-Learning is employed to dynamically optimize routing decisions based on real-time network states, thus facilitating adaptive and scalable congestion management. The Q-Learning algorithm updates its policy to favor paths that minimize future expected congestion, based on a reward structure tuned for network performance metrics such as throughput and packet loss. The impacts of this integrated approach are profound, improving throughput and reducing latency significantly compared to conventional methods. Through rigorous performance evaluations, the proposed method not only demonstrates superior scalability and robustness across diverse network conditions but also shows remarkable improvements in network stability and service quality. This model presents a significant step forward in the design of adaptive, efficient, and robust congestion control mechanisms for wireless networks, marking a pivotal advancement in network management technology.

Article Details

Section
Articles