AirNet

Abstract

Drones, or Unmanned Aerial Vehicles (UAVs) are gaining a great deal of attention not just from military sector, but also from many different domains including the research community and industry. Recently multi-drone systems are receiving attention as they are more capable than the single-drone case. To realize a multi-drone system, one needs to develop a networking scheme to connect drones and a control algorithm to smartly coordinate each drone's movement. Since it is expensive and time-consuming to conduct drone experiments, simulation is the primary choice for developing a drone system. There exists a number of network simulators and drone simulators, but no simulators are capable of simulating both networking and drones' dynamics at the same time.

We introduce AirNet, a joint simulator for networking and drones. AirNet is developed by integrating a drone simulator AirSim and a network simulator OMNeT++. As an extension of AirNet, we present AirNet-gym which is a framework to develop multi-agent reinforcement learning environments for multi-drone systems.