An Ad-hoc Planner for the Mars Express Mission Martin Kolombo, Martin Pecka, Roman Bart´ ak Charles University, Faculty of Mathematics and Physics Malostransk´ e n´ am. 2/25, 118 00 Praha 1, Czech Republic {kolombomartin,peci.jh}@gmail.com, bartak@ktiml.mff.cuni.cz Abstract Complete planning and scheduling of all spacecraft op- erations is a challenging area with the remote agent ex- periment at Deep Space 1 being a pioneering system. Still the complete approach is rare in practice. For ex- ample, in the Mars Express (MEX) mission, planning and scheduling techniques are used to solve some sub- problems namely scheduling command upload and data download. In this paper we describe an approach to gen- erate a complete schedule of the spacecraft that includes planning and scheduling of science, command uplink, data downlink, maintenance, and pointing operations. The proposed solving approach was designed to plan operations on the Mars Express (MEX) mission and it was motivated by the MEX challenge at the Fourth In- ternational Competition on Knowledge Engineering for Planning and Scheduling. The method is based on in- cremental addition of operations to a partial schedule and modifying the time allocation of already scheduled operations to fit the newly added operation. Despite its simplicity, the method seems to perform very well on experimental data, though a deeper evaluation with real data is still necessary. The paper briefly describes the problem solved, the integrated planning and scheduling algorithm, and the initial experimental results. Introduction The Mars Express Mission (MEX) is a successful mission of the European Space Agency with the spacecraft orbit- ing around Mars and producing 2-3 Gbit of scientific data per day. It is also one of the successful examples of ap- plication of planning and scheduling techniques in space. MEXAR 2 (Cesta et al. 2007) and RAXEM (Rabenau et al. 2008) are two tools operational at ESA-ESOC. MEXAR 2 was develop to schedule Data Dumping activities while RAXEM schedules Data Uplink activities. However, these tasks (downlink and uplink) are just two types of tasks nec- essary to operate the spacecraft. Naturally, the core role of MEX consists of scientific (observation) tasks that are com- plemented by the command uplink and data downlink activi- ties. There are also maintenance activities necessary to keep the spacecraft in a good condition. Currently the complete planning process is realized through a collaborative problem solving process between the science team and the mission planning team. Two teams of human planners iteratively re- fine a plan of all activities of the mission. In 2012 this com- plex planning problem has been proposed as a challenge for the Fourth International Competition on Knowledge Engi- neering for Planning and Scheduling (ICKEPS 2012). The goal was to develop a system that takes the description of scientific operations together with operational constraints, ground station visibility, a spacecraft trajectory etc. as its input and generates a complete schedule of all the uplink, science, downlink, maintenance, and auxiliary operations. Notice in particular that the flight dynamics is not part of the solution as the spacecraft trajectory is given as the in- put. The schedule must respect all operational constraints and maximize the scientific outcome. In this paper we describe a solving approach that we de- veloped for the ICKEPS challenge. We see the problem mainly as a scheduling problem because we can generate (plan) the main activities to be scheduled (uplink, science, downlink, maintenance) in advance and only some auxil- iary activities (pointing) need to be inserted to make the schedule consistent. This is very similar to inserting setup activities in the manufacturing scheduling problem (Bart´ ak 2003), which is a problem with some but weak planning component, where the scheduling component prevails. Our approach is based on incremental extension of a partial plan/schedule to which we add the science activities with their supporting (uplink and downlink) activities. This is similar to the approach used in the Mars Exploration Rovers employed by MAPGEN (Bresina et al. 2005). Science ac- tivities to be scheduled are explored by backtracking search while the time allocation of activities is done using a form of local search. Some pre-processing and post-processing is used to add the maintenance and pointing activities. The paper is organized as follows. We will first shortly describe the MEX domain, the full description can be found in (Fratini and Policella 2012). Then we will introduce the core concept of our approach that is is based on maintaining a single timeline of activities. After that we will describe the details of the proposed scheduling algorithm including the pre-processing and post-processing stages. The paper will be concluded by the description of current user interface and summary of some initial experiments.