Design of an Integrated Model Using Hybrid PSO-ABC and DQN for Energy-Efficient Healthcare Sensor Networks

Abstract

Energy efficiency and reliability in communication is critically necessary for HSNs as they have stringent deployments in dynamic and resource-constrained environments. Traditional clustering and routing algorithms fail to achieve an acceptable trade-off between energy efficiency and network performance, especially in changing network conditions and topologies. All of the existing methods have proved to be satisfactory in specific scenarios but usually suffer from issues such as complexity in dealing with load balancing, fault tolerance, and maintaining low latency, leading to lower lifetime for networks and degraded Quality of Service. In this paper, we present a multiple objective optimization framework to handle the above issues. This framework integrates the Hybrid Particle Swarm Optimization and Artificial Bee Colony algorithms along with Reinforcement Learning-based Dynamic Clustering with Deep Q-Networks and the Genetic Algorithm-enhanced hierarchical clustering model. This hybrid PSO-ABC method will exploit PSO's global exploration capabilities combined with local search refinement of ABC for optimizing the clustering and routing path of the nodes to eventually utilize much energy efficiency and delivery packet rates. Meanwhile, the RLDC using DQN dynamically and adaptively changes the structure of the clustering with the runtime status to bring optimization to the routing policies into fault tolerance and lower latency. Finally, GA-based technique ensures optimal cluster head selection and energy-efficient inter-cluster communication through evolutionary optimization techniques. Extensive simulations showed that the proposed framework outperformed existing approaches with a 15% performance improvement in terms of energy efficiency, 10% improvement in network throughput, and a 9% increase in packet delivery ratio along with enhancements in fault tolerance and network lifespan. Therefore, the results show that intelligent hybrid optimization succeeds in meeting the challenging demands of future HSNs.