Application of Deep Learning in Autonomous Control of Unmanned Aerial Vehicles (UAVs) in Mobile, Wireless Networks

Recent developments in robotics, miniaturisation, sensors, and communications technology have revolutionised Unmanned Aerial Vehicles (UAVs, also known as drones). Due to their deployment flexibility, low cost, and programmable control features, UAVs are being considered in a wide range of application domains including mobile, and wireless communications in which UAVs act as Flying Base Stations. Due to the dynamic and complex nature of the network requirements, e.g., change of traffic load and user concentrations, mobility of the UAVs needs autonomous control that can satisfy these requirements. The network scenario incorporates dynamic settings, especially when multiple flying UAVs are covering a large area, their changing energy levels, the current locations of the UAVs, types of services that they can provide, the changes in traffic load demands, etc., it is not a trivial task to determine a-priori the areas that will need more resources by the UAVs to plan their movements. To this end, Deep Learning (DL) technologies will be employed to train a Neural Network agent which will be able to make decisions in an online fashion to satisfy such networking demands. Based on the specific requirements, MATLAB or TensorFlow will be used to develop and train the DL models.