Robust machine learning-enabled routing for highly mobile vehicular networks with PARRoT in ns-3

Abstract

As implied by the high grade of relative mobility, the inherent network topology dynamics render aerial and ground-based vehicular mesh routing a highly challenging task. Since existing protocols are often not able to timely adopt their decision-making to the actual network conditions, they fail to provide reliable and efficient data delivery mechanisms. In this paper, we present the ns-3 integration of Predictive Ad-hoc Routing fueled by Reinforcement learning and Trajectory knowledge (PARRoT), a novel reinforcement learning-enabled routing protocol that integrates knowledge about the future motion of the mobile agents into the routing process.