Development of a simulation environment for AI algorithms applied to space missions

Space missions are becoming more complex due to the increased sensitivity and capability of onboard instruments and the need to use multiple spacecraft to accomplish more challenging tasks. To meet this complexity, current ground and space operations will need to use new paradigms to implement these missions while keeping costs affordable. Automation will play a fundamental role in that sense by reducing the costs and increasing the efficiency and performance of spacecraft. In this context, the Italian start-up company AIKO is developing state–of–the–art AI algorithms for space applications where the most recent technologies in deep learning, expert systems, and intelligent agents are used to process spacecraft data generated during a mission in order to take decisions autonomously without relying on ground station operations. However, successfully applying AI in those fields requires powerful and realistic virtual environments that enable risk–free training and testing of learning agents. This thesis presents the work carried out during a six months internship at AIKO aimed at developing the GOO4S (Googleplex Starthinker Space Systems Simulator) application, a simulation environment for space missions capable of simulating a rich spectrum of spacecraft behaviours and interacting with learning agents during the simulation. GOO4S is an application developed mostly in Python using the Object-Oriented Programming paradigm, with which a user or an external application can easily interact by posting commands and receiving relevant simulation data. The work carried out during the internship was mainly focused on the development of the Astrodynamic, Platform, and Environment modules of GOO4S. Firstly, the Astrodynamic Module has been implemented. It is used to simulate the spacecraft dynamics considering the most relevant perturbations affecting spacecraft’s motion and to detect the occurrence of several events such as spacecraft–to–ground stations visibility and Sun occultations. To monitor the spacecraft’s resource consumption and availability during a mission, the Platform Module has been implemented. It provides a high–level simulation of the spacecraft’s subsystems, in particular the Electrical and Power Subsystem (EPS), the Command and Data Handling Subsystem (CDHS), and the Payload component. Finally, the Environment Module has been implemented in order to simulate the acquisition of Earth–Observation data such as high–resolution optical images, atmospheric and temperature data.