High-Cohesion Multi-Objective Optimized Deep Neural Network for Energy Efficient and Load Balanced Environmental Data Transmission in WSN

Abstract

A Wireless Sensor Network (WSN) is a network comprising a large number of independent sensors deployed across various locations to sense and collect environmental data. These sensors communicate wirelessly and are used in a wide range of applications, including environmental monitoring, agriculture, healthcare, smart cities, industrial automation, and military systems. Environmental analytics applications include monitoring air and water quality, tracking climate change, analyzing human activities, and more. In such contexts, WSNs play a crucial role by enabling efficient data collection, processing, and transmission to provide valuable information from remote or inaccessible areas. During data transmission, energy optimization in WSNs is essential to increase the network’s lifetime and enhance the operational time of sensor nodes. The conventional methods have been developed for energy-efficient data transmission, especially regarding delay-aware and load-balanced transmission remain challenges. To address these challenges, a novel method called High-Cohesion Multi-objective Optimized Deep Neural Network (HiMO-DNN) is proposed. This method ensures energy-efficient environmental data communication with minimal latency and high data delivery rates. The HiMO-DNN model incorporates optimization and regression techniques to achieve energy-efficient load balancing in WSNs. The main aim of the HiMO-DNN is a Deep Neural Network, which consists of four layers namely an input layer, two hidden layers, and an output layer. The input layer comprises a varying number of sensor nodes. First, the HiMO-DNN model evaluates the residual energy of the sensor nodes. Based on this energy measurement, the nodes are grouped using a high-cohesion correlation clustering technique. A cluster head is then selected for each cluster to enhance the efficiency of data transmission. Next, a Multi-objective Stochastic Sampled Crow Search Optimization algorithm is employed to identify the nearest neighbor cluster head with the less loaded, thereby improving data transmission throughput while minimizing latency. Finally, the output layer observes the results of resource-efficient and load-balanced environmental data transmission. Experimental analysis is conducted using metrics such as including energy efficiency, Transmission Success Ratio, transmission latency, throughput, and Data drop rate, evaluated across varying amounts of environmental data and different numbers of sensor nodes.