Optimization of cost and capacity of broadband satellite system and resources management using Machine Learning techniques

Very High Throughput Satellite (VHTS) systems have an important role to play as a complement to future 5G terrestrial networks to meet the growing traffic demand. In the near future, VHTS systems are expected to reach a transmission capacity of 1 Tbps based on frequency reuse/polarization and multibeam coverage schemes. However, the traffic demand in the service area is not uniform and is also changing throughout the day. This means that with a traditional payload, some beams have insufficient resources and others have wasted resources. One solution to this problem is flexible payloads that allow satellite resources to be modified according to traffic demand. According to operators, the main challenges in Satellite Communications (SatComs) is to achieve new generation VHTS systems capable of satisfying traffic demand and to know how to manage resources in an optimal and autonomous way, thus emerging the problem of Dynamic Resource Management (DRM). With this in mind, this thesis studies the optimization for the design of new generation VHTS systems. The study is divided into two parts, satellites with fixed payload and satellites with flexible payload. For the first part, an optimization method is developed that minimizes the cost per Gbps in orbit and maximizes the capacity per beam, as a function of the number of beams, user G/T and annual availability. As an intermediate step between flexibility and a fixed system, the possibility of having a payload that provides coverage with irregularly sized beams depending on traffic demand is studied. While, for flexible systems, new optimization techniques belonging to Machine Learning are studied to manage resources dynamically and autonomously in the system. The results of this thesis provide new contributions for the design of new generations of VHTS broadband satellites and open a possibility for new research lines