Ant Colony Optimization applied to the Planning of a Data Relay Space Mission Evridiki V. Ntagiou 1 , Nicola Policella 3 , Claudio Iacopino 2 , Roberto Armellin 1 , Alessandro Donati 3 1 Surrey Space Centre, University of Surrey, Guildford, UK Email: {e.ntagiou, r.armellin}@surrey.ac.uk 2 Surrey Satellite Technology Ltd, Guildford, UK Email: c.iacopino@sstl.co.uk 3 ESOC, European Space Agency, Darmstadt, Germany Email: {nicola.policella, alessandro.donati}@esa.int Abstract Data relay space missions are becoming more popular as the amount of data produced on board LEO spacecraft has greatly increased, along with the size of the user community that access them. Planning those missions requires solving an oversubscribed scheduling problem. In this paper we propose the application of a Swarm Intelligence algorithm to the design of the planning system of a Data Relay space mission. We discuss characteristics of the system like the quality of the solutions produced, its reliability and adaptability, and provide some preliminary results. Introduction Missions observing and monitoring the Earth have received an increasing amount of attention due to the importance of the data they provide us with, ranging from weather forecast, science applications to natural disaster data. Both the population of Earth Observation (EO) spacecraft and users that can access their data have increased. The growth in the number of EO spacecraft results in a corresponding increase of the amount of data produced on board. The data needs to be transmitted to the ground as soon as possible in order to free on board resources, or satisfy an urgent user request. To that respect, data relay missions are being designed and operated. In these missions, Geosynchronous (GEO) satellites act as relays of payload data among Low Earth Orbit (LEO) spacecraft, Ground Stations and possibly aircraft. As a result, LEO spacecraft can communicate with Ground Stations for data downlink and command uplink with reduced time delays in the transmission, no matter what Copyright (c) 2017 All rights reserved. their relative position is. NASA TDRS (Gramling & Chrissotimos, 2008) and ESA EDRS (Wallrapp, Ballweg, & Gataullin, 2011) are two examples of such missions. Data Relay mission planning can be seen as an oversubscribed scheduling problem. In this family of problems, the initial set of requests is larger than the set of the tasks that can be allocated, regardless of the planning horizon. Many approaches that can be found in the literature are permutation-based, i.e. the search space consists of different orderings of the requests based on some predefined priority. In particular, the following two steps are iterated: (1) a schedule is produced considering the current ordering and (2) the schedule is then evaluated and changes of the ordering are imposed to explore other areas of the search space. The first step usually involves a greedy scheduler that schedules the tasks based on the initial ordering, at the earliest possible time. Then, a meta- heuristic evaluates the solution produced and decides on an order permutation, which the scheduler turns into a solution anew. In (Barbulescu, Whitley, & Howe, 2004) the question whether making big changes results in better solutions than making small changes is discussed. A genetic algorithm, Genitor (Whitley, 1989) and Squeaky Wheel Optimization (SWO) (Joslin & Clements, 1999) outperform local search methods, specifically because of the big permutation changes that they offer in a search space that is found to have many plateaus. An algorithm resulting from a combination of these methods is described in (Barbulescu et al., 2006) that offers very good results. In this paper we suggest a different approach based on Swarm Intelligence. The idea is to try to balance the exploration and exploitation phases, which is a form of balancing, at each step, between making big changes to a schedule and making small ones. The search space is not