A comparative analysis of algorithms for satellite operations scheduling

Scheduling is employed in everyday life, ranging from meetings to manufacturing and operations among other activities. One instance of scheduling in a complex real-life setting is space mission operations scheduling, i.e. instructing a satellite to perform fitting tasks during predefined time periods with a varied frequency to achieve its mission goals. Mission operations scheduling is pivotal to the success of any space mission, choreographing every task carefully, accounting for technological and environmental limitations and constraints along with mission goals.;It remains standard practice to this day, to generate operations schedules manually ,i.e. to collect requirements from individual stakeholders, collate them into a timeline, compare against feasibility and available satellite resources, and find potential conflicts. Conflict resolution is done by hand, checked by a simulator and uplinked to the satellite weekly. This process is time consuming, bears risks and can be considered sub-optimal.;A pertinent question arises: can we automate the process of satellite mission operations scheduling? And if we can, what method should be used to generate the schedules? In an attempt to address this question, a comparison of algorithms was deemed suitable in order to explore their suitability for this particular application.;The problem of mission operations scheduling was initially studied through literature and numerous interviews with experts. A framework was developed to approximate a generic Low Earth Orbit satellite, its environment and its mission requirements. Optimisation algorithms were chosen from different categories such as single-point stochastic without memory (Simulated Annealing, Random Search), multi-point stochastic with memory (Genetic Algorithm, Ant Colony System, Differential Evolution) and were run both with and without Local Search.;The aforementioned algorithmic set was initially tuned using a single 89-minute Low Earth Orbit of a scientific mission to Mars. It was then applied to scheduling operations during one high altitude Low Earth Orbit (2.4hrs) of an experimental mission.;It was then applied to a realistic test-case inspired by the European Space Agency PROBA-2 mission, comprising a 1 day schedule and subsequently a 7 day schedule - equal to a Short Term Plan as defined by the European Space Agency.;The schedule fitness - corresponding to the Hamming distance between mission requirements and generated schedule - are presented along with the execution time of each run. Algorithmic performance is discussed and put at the disposal of mission operations experts for consideration.Scheduling is employed in everyday life, ranging from meetings to manufacturing and operations among other activities. One instance of scheduling in a complex real-life setting is space mission operations scheduling, i.e. instructing a satellite to perform fitting tasks during predefined time periods with a varied frequency to achieve its mission goals. Mission operations scheduling is pivotal to the success of any space mission, choreographing every task carefully, accounting for technological and environmental limitations and constraints along with mission goals.;It remains standard practice to this day, to generate operations schedules manually ,i.e. to collect requirements from individual stakeholders, collate them into a timeline, compare against feasibility and available satellite resources, and find potential conflicts. Conflict resolution is done by hand, checked by a simulator and uplinked to the satellite weekly. This process is time consuming, bears risks and can be considered sub-optimal.;A pertinent question arises: can we automate the process of satellite mission operations scheduling? And if we can, what method should be used to generate the schedules? In an attempt to address this question, a comparison of algorithms was deemed suitable in order to explore their suitability for this particular application.;The problem of mission operations scheduling was initially studied through literature and numerous interviews with experts. A framework was developed to approximate a generic Low Earth Orbit satellite, its environment and its mission requirements. Optimisation algorithms were chosen from different categories such as single-point stochastic without memory (Simulated Annealing, Random Search), multi-point stochastic with memory (Genetic Algorithm, Ant Colony System, Differential Evolution) and were run both with and without Local Search.;The aforementioned algorithmic set was initially tuned using a single 89-minute Low Earth Orbit of a scientific mission to Mars. It was then applied to scheduling operations during one high altitude Low Earth Orbit (2.4hrs) of an experimental mission.;It was then applied to a realistic test-case inspired by the European Space Agency PROBA-2 mission, comprising a 1 day schedule and subsequently a 7 day schedule - equal to a Short Term Plan as defined by the European Space Agency.;The schedule fitness - corresponding to the Hamming distance between mission requirements and generated schedule - are presented along with the execution time of each run. Algorithmic performance is discussed and put at the disposal of mission operations experts for consideration