Preferred Solutions of the Ground Station Scheduling Problem using NSGA-III with Weighted Reference Points Selection Margarita Antoniou Computer Systems Department Joˇ zef Stefan Institute Joˇ zef Stefan Intl. Postgraduate School Ljubljana, Slovenia margarita.antoniou@ijs.si Gaˇ sper Petelin Computer Systems Department Joˇ zef Stefan Institute Joˇ zef Stefan Intl. Postgraduate School Ljubljana, Slovenia gasper.petelin@ijs.si Gregor Papa Computer Systems Department Joˇ zef Stefan Institute Joˇ zef Stefan Intl. Postgraduate School Ljubljana, Slovenia gregor.papa@ijs.si Abstract—The Ground station Scheduling Problem refers to the allocation of the communication tasks, among the ground stations and satellites. In general, the problem is formulated as a many-objective problem. The NSGA-III is an algorithm developed to solve such problems. Due to its selection operator that uses a number of reference points, the NSGA-III gives users the option to specify their own reference points. In this paper, we use this opportunity by generating distributed reference points, shifted depending on weights. A specific weight is assigned to each objective that corresponds to reference points for preferred solutions. The generation of these reference points and their effect on the final objective function values of the Pareto front are first tested on the DTLZ2 test function with 4 objectives and for a different combination of weights. Finally, various weights are applied to an instance of the Ground station Scheduling Problem, leading to Pareto fronts that favor specific objectives and feasible schedules. Index Terms—many-objective, satellite scheduling problem, NSGA-III, preferred solutions I. I NTRODUCTION The development of space science and technology has increased the number of satellites orbiting the earth. At the same time, the network of ground stations available to commu- nicate with these satellites remains rather limited. The service of satellites relies on this communication between ground stations and satellites. Therefore, an appropriate allocation of the time for satellites communicating ground stations is very important for the space industry. This gives rise to particularly challenging scheduling problems as the resources between space and ground entities are limited. The Ground station Scheduling Problem (GSP) involves reasonable arrangements of satellites, time windows for com- pleting tasks, e.g., telemetry tasks, and maximizing the ground This work was supported by the Slovenian Research Agency (research core funding No. P2-0098), by the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 692286 (SYNERGY), and under the Marie Sklodowska-Curie grant agreement No. 722734 (UTOPIAE). station usage. The GSP is a many-objective problem in its general formulation [1]. It has been solved among others by weighted-sum methods [2], multi-objective approaches such as with a memetic algorithm [3] or genetic algorithm [4], an ant colony optimization [5] and heuristic and local search methods [6]. These methods find either one solution of the Pareto front or several well-spread solutions of the Pareto front. The optimization methods for multi/many-objective op- timization problems are divided into 3 categories from a decision-making point of view. When no prior information about the preferred solutions exist, the whole Pareto front needs to be explored, and the decision-maker is choosing one or more preferred optima from the Pareto-optimal solution set at the end of the optimization (a posterior method), which is basically any traditional multi/many-objective evolutionary algorithms, such as NSGA-II [7], MOEA/D [8], etc. When some information regarding the preferred region of a Pareto Front is known in advance, the decision-makers are usually interested in exploring only that small part of the front and a priori optimization methods can be applied, such as [9]. Lately, interactive preference methods have been proposed, e.g. in [10], where the decision maker’s preferences are taken into account dynamically during the optimization, saving computational cost. The formulation of GSP we adopt in this paper is the one used in [2], considering the following objectives: 1) maximizing the events that fall inside the available Access Windows of the ground stations, 2) minimizing the clashes when more than one satellite is communicating with the same ground station, 3) maximizing the time that is required to finalize specific telemetry tasks, such as the download of images, and finally, 4) taking advantage as much as possible from the ground station network by minimizing their idle time. As we have already noted in [11] and [12], the objectives of access windows and clashes are in fact constraints. By simultaneously solving this many-objective problem, it may lead to solutions of the Pareto front, that result in infeasible final schedules. Therefore one is interested in a particular part 978-1-7281-8393-0/21/$31.00 ©2021 IEEE 1840 2021 IEEE Congress on Evolutionary Computation (CEC) 2021 IEEE Congress on Evolutionary Computation (CEC) | 978-1-7281-8393-0/21/$31.00 ©2021 IEEE | DOI: 10.1109/CEC45853.2021.9504886 Authorized licensed use limited to: Institute Jozef Stefan. Downloaded on September 15,2021 at 07:15:31 UTC from IEEE Xplore. Restrictions apply.