QMRNB: Design of an Efficient Q-Learning Model to Improve Routing Efficiency of UAV Networks via Bioinspired Optimizations

Abstract

– The design of efficient routing strategies for Unmanned Aerial Vehicle (UAV) Networks is a multidomain task that involves analysis of node-level & network-level parameters, and mapping them with communication & contextual conditions. Existing path planning optimization models either showcase higher complexity or cannot be scaled for larger network scenarios. Moreover, the efficiency of these models also reduces w.r.t. the number of communication requests, which limits their scalability levels. To get a better result over these challenges, this article provides an idea to design an efficient Q-Learning model to improve the routing efficiency of UAV networks via bioinspired optimizations. The model initially collects temporal routing performance data samples for individual nodes and uses them to form coarse routes via Q-Learning optimizations. These routes are further processed via a Mayfly Optimization (MO) Model, which assists in the selection of optimal routing paths for high Quality of Service (QoS) even under large-scale routing requests. The MO Model can identify alternate paths via the evaluation of a high-density routing fitness function that assists the router in case the selected paths are occupied during current routing requests. This assists in improving temporal routing performance even under dense network conditions. Due to these optimizations, the model is capable of reducing the routing delay by 8.5%, improving energy efficiency by 4.9%, and reducing the routing jitter by 3.5% when compared with existing routing techniques by taking similar routing conditions.