A Novel Multi-hop WSN Routing Approach with GOA and LSTM Neural Network

Abstract

Energy resource management is a challenging task in WSNs owing to very limited energy capacity in sensor nodes. This paper presents an effective multi-hop routing approach in WSN using the strengths of both GOA and an LSTM neural network so that WSN can be energy-efficient and may last long. The proposed algorithm works from the two major steps' perspective: GOA-based clustering and cluster head selection, while routing of cluster heads with the LSTM network. First of all, K-means clustering is applied to determine preliminary positions of cluster heads in order to enhance convergence speed and avoid trapping into suboptimal solutions. In this work, a new cost function was developed related to the clustering phase, taking energy consumption and distribution into consideration, which enhances network flexibility in further rounds. In the routing phase, a new mechanism will be based on the minimum spanning tree approach, avoiding that in any round, the LSTM network ends up with isolated or meshed WSN. Concretely, this LSTM network predicts the weights optimally for the minimum spanning tree algorithm to efficiently route within the fully connected graph of cluster heads. More importantly, the proposed LSTM is trained with data obtained using the GOA-PSO approach. Extensive simulations were conducted to investigate the performance of the proposed method for three scenarios where base station placements varied. These results depict that GOA-LSTM has outperformed the traditional LEACH protocol significantly in network lifetime by an average improvement of 155.2691% considering different locations. GOA-LSTM also depicts a slight lower performance compared to GOA-PSO but with significant computational savings, which makes GOA-LSTM more practical and viable regarding real-time applications. This work investigates a robust, energy-efficient routing approach that easily adapts to varied network topologies and operational conditions, thus prolonging the operation lifetime of WSNs.