Bio-inspired based Random Forest Approach for Depth-based Routing in Underwater Sensor Networks

Underwater Sensor Networks (USNs) face significant challenges such as limited bandwidth, high propagation delay, and energy constraints, which hinder reliable and energy-efficient data transmission—especially in deep and dynamic aquatic environments. Traditional routing methods often fail to adapt effectively when the sink node is mobile or positioned at varying depths. To address this problem, we propose a Bio-inspired Random Forest (Bio-RF) routing algorithm that integrates depth-aware decision-making to optimize the selection of forwarding nodes. The proposed model leverages machine learning principles inspired by natural selection, using features such as node depth, hop count, and residual energy to determine the optimal next-hop for packet forwarding. Among the evaluated variants, the Hippopotamus-Inspired RF model achieved the lowest Mean Squared Error (MSE) and minimized training time, demonstrating its suitability for learning accuracy. In contrast, the Kookaburra- and Firefly-Inspired RF models delivered faster inference, making them better suited for real-time routing decisions in time-critical applications. These results confirm the adaptability and intelligence of the proposed models over traditional depth-based protocols.