| Popis: |
Drones, or Unmanned Aerial Vehicles (UAVs), are a key part of many applications today such as e-commerce, agriculture, health care, and disaster management. The nodes of these networks are generally called Flying Ad-hoc Networks (FANETs), a subdomain of Mobile Ad hoc Networks (MANETs). These nodes have several challenges, including mobility management, rapid mobility, communication issues, dynamic topologies, energy efficiency, limited bandwidth, etc., which create a challenging environment for mobile ad-hoc networking. In this research, the proposed technique specifically addresses the challenges related to mobility management, rapid mobility, and communication issues arising from high mobility in nodes within FANETs. We present a novel approach where flying nodes are modelled using semi-Markov decision problems (SMDPs) based on state transitions and actions which are then solved using the Reinforcement Learning (RL) approach for reliable and effective communication in FANETs. We then test and validate the RL model by redesigning the Gauss Markov mobility model by adding two new parameters that enhance its performance. The flying node parameters were then emulated and analyzed based on the test results. The results showed improved node connectivity, mobility, and Quality of Service (QoS) in high mobile scenarios, such as a reduction in packet loss by 23%, a reduction in delay, and a slight improvement in throughput. Moreover, the delay factor has been improved by at least 30% over the existing results. |
