The identification and pursuit of Grand Challenges has been a hallmark of the highperformance computing arena for well over a decade. In recent years, many other technical communities, including the modeling and simulation (M&S) community, have begun defining Grand Challenge problems for their disciplines. While Grand Challenges themselves provide a useful focal point for research and development activities within a discipline, perhaps more important is the community dialogue that surrounds the formulation of Grand Challenge problems.

Within the M&S community, the dialogue surrounding the notion of Grand Challenges began with the First International Conference on Grand Challenges for Modeling and Simulation, which was held 27-31 January 2002 in San Antonio, TX, USA as part of Society for Computer Simulation (SCS) 2002 Western Multiconference. The conference program consisted of 15 papers and a panel.

The Dagstuhl seminar on Grand Challenges for M&S was dedicated to continuing this dialogue, with the goal of condensing ideas into a set of Grand Challenge problem statements that might serve to guide strategic research initiatives in modeling and simulation for the next decade.

The seminar was structured around various application and methodological areas of modeling and simulation:

• Simulation of cellular systems
• Simulation of air traffic
• Simulation large scale computer networks
• Simulation as part of agent-oriented software engineering
• Simulation in virtual manufacturing
• Simulation in military applications
• Parallel and distributed simulation
• Modeling and simulation methods

While the groups had unique perspectives derived from their particular application domain, they also shared a commonality derived from the modeling and simulation life cycle (i.e. understand a system, represent a system as a model, execute the model, analyze the results). Cognitive models of human actors, their decision processes, and their behavior are important in military applications, and in testing autonomous agent software. However, cognition processes are still little understood and "of the shelf" cognitive models that can be re-used in different settings do not exist. The same is true if we are looking at cellular, biological systems. The successful completion of the ambitious endeavor of the human genom project depends to a large degree on a better understanding of the behavior of cellular systems.

In dealing with complex systems, like cellular or cognitive systems, modeling and simulation has often played a role to support the development of theories and understanding of systems rather than predicting the systems' behavior. Efforts of the application area have to be combined with developing simulation systems that support an explorative approach to modeling and simulation more effectively. Whereas many techniques, e.g. hierarchical decomposition, object-oriented modeling and programming, graphical depiction of system behavior, visual modeling and programming, or agentbased modeling, have enhanced our ability to build and use complex models, despite efforts like HLA, still the challenge of re-usability of models seems largely unresolved, particularly if we are approaching the realm of multi-paradigm, multi-resolution modeling. Supporting multi-paradigm, multi-resolution modeling is arguably a central prerequisite to significantly advancing modeling and simulation in such diverse application areas like manufacturing, military, air traffic, biology, software development, and networks. Complex systems, e.g. the world wide web, do not only require new techniques for a more effective representation of systems. The efficient execution of these models poses unsolved problems as well. New parallel distributed simulation methods are needed not only to support an efficient simulation but to adapt themselves flexibly to the changing demands of a multi-resolution and multi-paradigm modeling.

During the seminar, a set of Grand Challenge problems statements from each of the application areas was formulated, and in some cases, possibilities for research agendas were sketched. While the results of the seminar offer a good starting point, and illustrate a number of intersections of interest across M&S application domains, more thought and effort is required to develop concrete research agendas in the multi-disciplinary arena of modeling and simulation.

Organization

Dagstuhl is dedicated to working groups. In contrast to traditional conference settings, the schedule offered plenty of time for working groups, discussions, and spontaneous activities. The week was divided into two parts (1-4, and 5-8 respectively) and allowed everybody to participate in two working groups during the seminar. To give an overview about the different areas, state-of-the-art plenary talks were given. Short presentations provided the opportunity for each participant to present his or her work, and ideas on Grand Challenges for Modeling and Simulation before the parallel working groups started. In plenary sessions the results of the working groups were presented. Intertwining working groups and plenary sessions helped to work on concrete challenges in the different groups and to support a cross fertilization among them. The seminar was a truly interdisciplinary event and all participants played an active role in driving the progress and content of the workshop.

As always, Schloss Dagstuhl and its ambiance, its unusual blend of the old with the new, the organization, and the very helpful staff contributed largely to the success of the seminar.