681 Deploying Interactive Mission Planning Tools Experiences and Lessons Learned Amedeo Cesta*, Gabriella Cortellessa*, Simone Fratini*, Angelo Oddi*, and Giulio Bernardi* *ISTC-CNR, Institute of Cognitive Science and Technology, National Research Council of Italy Via S. Martino della Battaglia, 44, I-00185 Rome, Italy e-mail: name.surname@istc.cnr.it Abstract This article contains a retrospective overview of con- nected work performed for the European Space Agency (ESA) over a span of 10 years. We have been creating and rening an AI approach to problem solving and injected a series of deployed planning and scheduling systems which have innovated agency’s mission planning practice. Goal of the paper is to identify general lessons learned and guidelines for work practice of the future. Specically, the work dwells on issue related to some key points that have contributed to strengthen the effectiveness of our ap- proach: the attention to domain modeling, the constraint- based algorithm synthesis, and the development of an end- to-end methodology to eld applications. Desirable fea- tures of space applications useful for protable and suc- cessful deployment on different ground segment opera- tions are also discussed. 1 Introduction Our research group has developed a continuous col- laboration with the European Space Agency (ESA) over a span of ten years. Topic of this collaboration has been the injection of ideas from the AI planning and schedul- ing research area to support operations of real missions. As summarized in Figure 1 we have performed activities in different directions: – In “product driven” activities we have developed solutions for specic mission problems. For ex- ample in a fruitful collaboration with the Mission Planning Team of the Mars-Express mission at ESOC we have addressed specic problems con- nected with data/command management from/to the remote spacecraft. This effort resulted in two oper- ational prototypes currently in daily use at mission control center at ESA. In particular Mexar2 supports the downlinking problem of the Mars-Express mem- ory [7], while Raxem [5] and its strengthened ver- sion Raxem2 [3], resolve the complementary prob- lem of the uplink of operational commands to the probe. Such an approach inevitably entails a huge implementation effort in terms of development: spec- ication extraction, design, implementation, mainte- nance. Figure 1. A summary timeline showing the work of Planning and Scheduling Team for the European Space Agency – In “process driven” activities we have attempted at developing general purpose tools for facilitating the design and synthesis of new products. In par- ticular within a larger project that spanned from 2007 to 2009, called the Advanced Planning and Scheduling Initiative (APSI) we have developed the Timeline-base Representation Framework (Apsi- Trf) [9]. APSI aim was to develop a general frame- work to improve the cost-effectiveness and exibil- ity of the mission planning systems (MPS) develop- ment. The tool offers a Java plaform with primitives to capture the specicity of an application domain and a given problem, thus fostering rapid and fast prototyping. An example of such support is repre- sented by the MrSPOCK a operational product for Mars-Express long term planning. A new example of “process driven” approach is the GOAC project an ongoing activity for designing future generation of robotic controllers [20]. This paper aims at presenting a comprehensive picture on these activities and an attempt at drawing general points that emerge from the work. In revising our collabora- tion with ESA we focus also on the open challenges and i-SAIRAS 2010 August 29-September 1, 2010, Sapporo, Japan