Energy efficient multi sink routing for large scale healthcare Internet of Things wireless sensor networks using reinforcement learning and multi objective optimization

Increasing the network lifetime of healthcare Internet of Things systems deployed in large-scale wireless sensor networks requires energy-efficient, reliable, and secure communications. However, existing approaches do not adequately support shared-path reuse, intelligent forwarding, node fault handling, and adaptive relay deployment, resulting in increased energy consumption, communication overheads, and transmission delays. To address these limitations, this paper proposes an energy-efficient multi-sink routing framework based on the Steiner Routing with Route Memory and Path Cache Mechanism (SRRMPCM). Initially, sensor nodes (SNs), relay nodes (RNs), and sink nodes (SiNs) are initialized, and secure hash codes are generated using the Cauchy–Schwarz Divergence Hash-based Message Authentication Code. Subsequently, neighboring nodes exchanged hello packets, and faulty nodes were identified using a Square-Wave Spectral Normalization-based Deep Long Short-Term Memory model. Cluster heads are then selected from the fault-free nodes using the Mantis Shrimp Pareto-Front Optimization Algorithm, followed by density-based node clustering. Relay nodes are dynamically deployed using Sigmoid Energy Gate-based Reinforcement Learning, and capacity-aware relay assignment is performed using the Kuhn–Munkres Airy Algorithm. The sensed data are aggregated at the cluster heads and transmitted to multiple sink nodes through the proposed routing framework, followed by secure node authentication. Experimental results demonstrate that the proposed framework achieves a routing latency of 454 ms and a routing overhead of 7.5% for a network consisting of 100 nodes, demonstrating its effectiveness in improving energy-efficient and reliable data transmission for large-scale healthcare wireless sensor networks.