AQ-OLSR: ADAPTIVE ROUTING PROTOCOL FOR UAV AD HOC NETWORK Merged with DEEP Q-NETWORK (DQN) ALGORITHM
Ali H. Wheeb
Abstract
Unmanned Aerial Vehicles (UAVs) have emerged recently due to rapid improvements in wireless technology and low-cost equipment, advancement in networking communication techniques, and increased demand from various industries that seek to leverage aerial data to improve their business and operations. As such, UAVs have become highly prevalent for various civilian, commercial, and military uses over the past few years. UAVs form a UAV ad hoc network as they communicate and collaborate wirelessly. UAV ad hoc networks are frequently deployed in three dimensions, and the variability of mobility models is determined by the type of tasks they are performing. The high dynamic topology of UAV ad hoc networks makes the design of routing protocols difficult. Thus, this paper aims to propose an adaptive routing protocol for a more reliable UAV ad hoc network. Specifically, the research proposes an adaptive routing protocol based on Reinforcement learning (RL) and the Deep Q-Network (DQN) algorithm, known as the Adaptive Q-Network OLSR (AQ-OLSR). The proposed AQ-OLSR protocol can dynamically adjust the time interval of sending hello messages. UAV states such as link connection time, the remaining energy of the UAVs, and network performance metrics of UAV ad hoc network are considered as input states for the DQN algorithm, while the optimum interval of the hello message will be the output. The proposed AQ-OLSR protocol was implemented and evaluated using the ns3-gym simulator against ML-OLSR and energy efficient hello (EE-Hello-OLSR) protocols in varying UAV speed and density scenarios. Simulation results show the significant advantage of AQ-OLSR routing protocol between the range of 2.4% to 21.4% performance improvement for all scenarios and metrics. The overhead reduction provided by the AQ-OLSR protocol can be used to reduce further energy consumption, which is crucial in UAVs ad hoc networks missions.