Intelligent clustering and routing protocol for wireless sensor networks using quantum inspired Harris Hawk optimizer and deep reinforcement learning

Abstract

Dynamic changes in wireless sensor networks (WSNs) present significant challenges to clustering and routing protocols, particularly impacting energy efficiency and network lifetime. Existing protocols often fail to address the trade-off between energy conservation and optimal cluster based routing, especially in highly dynamic environments. This paper proposes an Intelligent Clustering and Routing protocol for WSNs, called ICR-HHODRL, which innovatively integrates the Quantum-inspired Harris Hawk Optimizer (QHHO) for clustering and Deep Reinforcement Learning (DRL) for routing, aiming to improve energy efficiency, improve network throughput, and maximize network lifetime. The protocol minimizes message overhead by dynamically selecting optimal cluster head (CH) and forming clusters using QHHO with a new fitness function that considers node’s residual energy, average distance to neighbor nodes, and distance to the base station (BS), ensuring a balanced distribution of energy and CHs. Furthermore, ICR-HHODRL leverages a Dueling Double Deep Q-network (D3QN) with priority experience replay (D3QN-PER) for adaptive learning of optimal routing policies, addressing dynamic network conditions and enhancing load balancing. Experiment results show that the proposed ICR-HHODRL protocol outperforms several state-of-the-art clustering and routing protocols. Specifically, network lifetime is improved by 14.85%, 18.46%, 15.17%, 17.69%, and 14.77, network throughput is increased by 5.7%, 7.29%, 5.69%, 10.02%, and 7.21%, and network energy consumption is reduced by 16.96%, 20.03%, 15.51%, 8.61%, and 18.8%, compared to WOAD3QN-RP, MRP-ICHI, QoSCRSI, HHO-UCRA, and WOAC-HHOR, respectively. These findings highlight the protocol’s potential to significantly advance the state of the art in dynamic WSNs and offer promising solutions for low-power, resource-constrained networks.