Title:Towards Machine Learning-Enabled Context Adaption for Reliable Aerial Mesh Routing

Abstract:In this paper, we present Context-Adaptive PARRoT (CA-PARRoT) as an extension of our previous work Predictive Ad-hoc Routing fueled by Reinforcement learning and Trajectory knowledge (PARRoT). Short-term effects, as occurring in urban surroundings, have shown to have a negative impact on the Reinforcement Learning (RL)-based routing process. Therefore, we add a timer-based compensation mechanism to the update process and introduce a hybrid Machine Learning (ML) approach to classify Radio Environment Prototypes (REPs) with a dedicated ML component and enable the protocol for autonomous context adaption. The performance of the novel protocol is evaluated in comprehensive network simulations considering different REPs and is compared to well-known established routing protocols for Mobile Ad-hoc Networks (MANETs). The results show, that CA-PARRoT is capable to compensate the challenges confronted with in different REPs and to improve its Key Performance Indicators (KPIs) up to 23% compared to PARRoT, and outperform established routing protocols by up to 50 %.