Development of a comprehensive simulation software for spacecraft missions

Abstract

The growth of satellite mission, especially CubeSats, in terms of complexity and capabilities has required the development of dedicated orbit simulation software for mission planning and analysis. This thesis presents the development and uses of a simulation software that will be used to aid in the design of spacecraft missions. The process of developing the software architecture is described in stages from software requirement analysis to test and verification of the final implementation. During a space mission, a spacecraft may be placed in a variety of orbits for different purposes. Preliminary mission design needs to consider all mission phases to meet the needs of more complex missions. To effectively design an orbit, it is important to clearly define the purpose of the orbit and regularly review and reassess this purpose as mission requirements evolve or become more defined. It is also important to consider alternative orbit designs, as there may be multiple options that are viable. For example, a single large satellite in a geosynchronous orbit or a group of smaller satellites in low-Earth orbit may both be effective for communication purposes. Multiple different designs are often compared to find the orbit that best accommodate the mission requirements. There are various criteria that have to be considered according to the mission, such as determining the communication links between satellites and ground stations, and finding the time intervals when there is a pass or eclipse, which allow determining the requirements of communication and power systems. For an Earth-observing satellite, the orbit that has the most revisit time for desired locations and properties of the optical system such as the field of view should be determined. The aim of this work is to make use of the software tools to create a simulation software that provides a framework for efficient analysis and planning of satellite missions that include earth observation, communication, and scientific objectives in order to helps the mission design process by giving the ability to make fast and reliable decisions regarding the satellite system requirements. The developed software implements multiple orbit propagators, with the most prominent being the High-precision Orbit Propagator (HPOP) which takes into account all of the forces that can be modelled so far. However, physics-based models alone are insufficient for accurately predicting orbits and avoiding collisions, as demonstrated by previous collisions caused by such predictions. Our knowledge of the physical world is not sufficient enough to create perfect models as it is near impossible to predict some perturbations precisely, such as solar activity which is only an approximation based on statistical data, as well as the atmosphere models and the area of the satellite that drag force affects, which also change depending on the attitude model. In order to improve the accuracy of these models, a machine-learning approach that utilizes the past flight data is proposed. Models of orbit prediction errors can be learned directly from a large amount of historical data, allowing for predictions without explicitly modeling forces or perturbations. Hence, a neural-networks model was trained and its impact was demonstrated. The software is developed using various programming languages. The user interface is programmed in JavaScript, using HTML and CSS. Orbital analyses and other computation heavy tasks were performed in C++ as it has the benefits of modular design, less resource use and fast execution speed, as well as good portability. Python was used for model training and artificial intelligence methods due to the enormous number of scientific libraries it includes. Electron framework was used to provide cross-platform compatibility. For real-time data visualization in both 2D and 3D, the Cesium framework was implemented using Bing as data provider for satellite imagery and terrain modelling. The simulation results show that the software succeeds in attaining high execution speed and precision. The results indicate that the proposed solution can be useful in reducing time and effort put into the mission design process as well as increase the rate of success for both Earth and interplanetary missions. The developed software can be used for real-world mission design and operations, as a tool for education and engineering studies, and public engagement. The structure of the thesis is as follows. First the mathematical background and celestial relationships that are widely used in mission planning are explained along with the algorithms used to implement them into the software. These include coordinate system transformations, various orbital elements, and time systems. Then orbital propagation is explained starting from two-body motion, which is the basis for all equations of motion, followed by adding perturbations and other forces acting on the satellite to facilitate the high-precision numerical propagator. Analytical propagators that are widely used are explained as well. Then, the design and implementation of a neural-networks model that is trained to improve the accuracy of the numerical propagators is described. The next chapter focuses on the simulation environment and the capabilities of the software. It is possible to easily create mission-specific orbits such as SSO, GEO, and Molniya, predict the visibility of satellites from different locations on the ground, determine the eclipse intervals, compute communication link budget, analyze on-orbit power generation, and perform basic maneuvers within the simulation environment. The software principles, architecture, development process, and user interface design is thoroughly explained as well. Finally, verifications using real data and satellite observations are performed and results are presented, which show that the developed software is ready for real mission use, and possible future developments are discussed.