Coordinate attention-driven multi-agent adaptive deep reinforcement learning model for resilient and energy-efficient routing in WSNs

Wireless Sensor Networks (WSNs) is considered as essential role in intelligent healthcare applications. The proposed model designs a novel deep-reinforcement learning model that leverages optimal CH selection to achieve sustainable and secure transmission of data in resource-constrained WSN environments. Initially, the intrusion detection process is performed to detect any type of intrusion in the WSN with the aid of an Adaptive Dense Elman Recurrent Neural Network (ADE-RNN). This proposed framework integrates optimal cluster head selection with intelligent routing mechanisms to enhance WSNs' performance in various domain-specific systems. Firstly, the CHs are chosen using the Synergistic Evolution-based Superb Fairy-wren Optimization Algorithm (SESFOA) that minimizes energy consumption, communication delay, and both intra- and inter-cluster distances. This ensures efficient data aggregation and transmission while preserving node longevity. Once clusters are formed, routing is managed through the developed Coordinate Attention-based Multi-Agent Adaptive Deep Reinforcement Learning (CAMA-ADRL) strategy. This enables each agent to learn and make decisions based on local observations and global feedback, while prioritizing the spatially significant nodes and paths. So, the performance of the path selection is enhanced, the developed technique undergoes parameter optimization using the SESFOA Optimization Algorithm.