SJO-MRL: Energy efficient path selection algorithm for WSNs

Wireless Sensor Network (WSN) comprises a group of sensors spread out in a specific area to observe environmental events and communicate data amongst themselves via wireless connections. The effective path selection in networks involves choosing optimal routes for data transmission considering parameters like distance, speed, energy efficiency, content of the data, terrestrial coordinates of sensor nodes, ordering of data, traffic, security, QoS parameters, past data current network conditions, etc. It is crucial for optimizing network performance by minimizing energy consumption and ensuring reliable data transmission. This research introduces a new model called Salar-Jelly Optimization based Markov Reinforcement Learning (SJO-MRL) for choosing paths in WSN. Implementing clustering techniques for data gathering is crucial in minimizing energy usage and prolonging the network's durability. The cluster head (CH) is nominated using the Hybrid Energy Efficient Distributed Clustering Protocol (HEED), considering various factors like node energy level, communication cost, the energy level of neighboring nodes, elect threshold, residual energy distribution, periodic re-clustering, randomized CH selection, and so on. The proposed algorithm uses the Markov Reinforcement Learning model to find optimal paths by learning behavior. Including sleep scheduling in data gathering can further encourage energy efficiency. A nature-inspired Hybrid Salar-jelly-based optimization combines Jellyfish and Salmon Optimization to mimic natural processes to schedule when nodes should be in an active (awake) or inactive (sleep) state. Offline model training optimizes computational resources. The results obtained show that at a specific time (99 ms), there is a very low delay of 0.9900 ms, and the data throughput is high at 499.89 bps. The energy used is also quite low, at 0.0319 J. Similarly, during training percentage (TP) 90, this model achieved a high accuracy rate of 95.73%, sensitivity of 96.42%, and specificity of 95.92%.