Publication

Abstract : Wireless Sensor Networks (WSNs) consist of low-power sensor nodes that communicate with a central hub to collect and transmit data. Since these nodes operate on limited battery power, energy efficiency is a critical concern. Traditional routing protocols often struggle with dynamic topology changes, heterogeneous device support, and scalability limitations. To address these challenges, this paper proposes an Energy-Efficient Clustering with Optimized Action Sequences-Based Q-Learning Algorithm for routing in heterogeneous WSNs. The proposed approach leverages Q-learning to determine optimal routing paths by selecting Cluster Heads (CHs) based on attribute-based clustering, while neighboring nodes form clusters. The CHs aggregate data from cluster members and forward it to the sink node. Additionally, the Adaptive Coati Optimization Algorithm (ACOA) enhances the action sequences of Q-learning, preventing isolated nodes and further optimizing routing efficiency. The proposed Q-Learning-based Routing Algorithm (QLRA) achieves optimal routing by dynamically adjusting Q-values based on selected action sequences. The performance of the approach is evaluated through extensive simulations conducted in MATLAB under three distinct scenarios: varying the number of rounds, the number of nodes, and the percentage of CH selection. The proposed Optimized QLRA with clustering and ACOA are compared against three baseline methods, such as Optimized QLRA without clustering, Traditional QLRA with clustering, and Traditional QLRA without clustering. The results demonstrate a significant 11.82% increase in throughput and a 25.19% reduction in energy consumption, confirming the proposed algorithm's effectiveness in improving energy efficiency, network lifetime, and data delivery reliability.