Energy-efficient adaptive routing protocol for EH-WSNs based on deep reinforcement learning and fuzzy clustering

Abstract In energy-harvesting wireless sensor networks (EH-WSNs), efficient energy management and reliable data transmission are crucial for prolonging network lifetime and improving performance. This paper presents an Adaptive Energy-Efficient Routing Protocol that integrates Deep Reinforcement Learning (DRL) and Fuzzy Clustering to address key challenges, such as energy consumption, network reliability, and data throughput. The proposed protocol leverages fuzzy logic to optimally select cluster heads based on energy levels, distance, and network density, thereby ensuring balanced energy usage among nodes. Additionally, DRL is employed to dynamically determine the best routing paths that minimize energy expenditure while maintaining reliable communication. Simulation results demonstrate that the proposed protocol outperforms existing approaches, such as Low Energy Adaptive Clustering Hierarchy (LEACH) and Efficient Routing Awareness Scheduling (ERAS), in terms of network lifetime, packet delivery ratio (PDR), throughput, and total energy consumption. Specifically, the proposed protocol achieves up to 25% longer network lifetime, 20-25% higher throughput, and 15-20% lower energy consumption compared to the benchmark protocols. These results highlight the effectiveness of combining fuzzy logic and DRL for adaptive routing in EH-WSNs, making the proposed solution highly suitable for real-world energy-constrained wireless sensor network applications.