Abstract

Wireless Sensor Networks (WSNs) are challenged by the need for optimized Energy Consumption (EC), efficient Data Aggregation (DA), and reliable routing due to their dynamic topologies and limited resources. Existing solutions like TEAMR and DDQNDA address these concerns but face significant drawbacks—TEAMR lacks adaptability to rapidly changing topologies, while DDQNDA suffers from high computational overhead and delayed convergence, hindering its effectiveness in real-time scenarios. To overcome these limitations, this paper introduces the Adaptive Reinforcement Learning (RL)-Based DA and Routing Optimization (ARL-DARO) algorithm. The proposed methodology follows a systematic approach, beginning with cluster formation and Cluster Head (CH) selection (CHS) using the Grey Wolf Optimizer (GWO), which ensures Energy-Efficient (EE) clustering and optimal CH selection. In the next step, trust factors such as Node Connectivity (NC), Residual Trust (RT), and Cooperation Rate (CR) are integrated into Quality of Service (QoS) metrics as part of the Fitness Function(FF) to enhance route reliability and security. Finally, the ARL-DARO algorithm is employed to dynamically optimize both data aggregation and routing. It leverages Q-learning to select optimal routes based on energy efficiency, security, and link reliability, further reducing data redundancy and improving adaptability to real-time network changes. Performance is assessed using parameters such EC, packet delivery ratio (PDR), end-to-end latency (E2E delay), throughput, and network lifetime (NL) across networks with 100, 200, 300, 400, and 500 nodes. Results show that ARL-DARO significantly reduces energy consumption by up to 45%, increases throughput by 30%, and extends network lifetime, proving its effectiveness over existing methods.