Science and engineering of large scale socio-technical simulations.

Computer simulation is a computational approach whereby global system properties are produced as dynamics by direct computation of interactions among representations of local system elements. A mathematical theory of simulation consists of an account of the formal properties of sequential evaluation and composition of interdependent local mappings. When certain local mappings and their interdependencies can be related to particular real world objects and interdependencies, it is common to compute the interactions to derive a symbolic model of the global system made up of the corresponding interdependent objects. The formal mathematical and computational account of the simulation provides a particular kind of theoretical explanation of the global system properties and, therefore, insight into how to engineer a complex system to exhibit those properties. This paper considers the methematical foundations and engineering princaples necessary for building large scale simulations of socio-technical systems. Examples of such systems are urban regional transportation systems, the national electrical power markets and grids, the world-wide Internet, vaccine design and deployment, theater war, etc. These systems are composed of large numbers of interacting human, physical and technological components. Some components adapt and learn, exhibit perception, interpretation, reasoning, deception, cooperation and noncooperation, and have economic motives as well as the usual physical properties of interaction. The systems themselves are large and the behavior of sociotechnical systems is tremendously complex. The state of affairs f o r these kinds of systems is characterized by very little satisfactory formal theory, a good decal of very specialized knowledge of subsystems, and a dependence on experience-based practitioners' art. However, these systems are vital and require policy, control, design, implementation and investment. Thus there is motivation to improve the ability to comprehend them by use of whatever means, including computer simulation. Moreover, the general theoretical understanding of the system properties provided by the formalization of simulation, is of great value given the otherwise poor state of understanding of the systems themselves. Theoretically, Sequential Dgnamical Systems (SDS) are introduced as a mathematical model of discrete simulation. Sequential dynamical systems are compositions of local maps. The order of composition reflects casual relationships between individual agents abstracted as functions and the locality of the functions reflects limited interaction and knowledge of the entire system available to each agent. The properties of SDS are very general and allow much deeper understanding of both the simulation and the simulated system. The last part considers the engineering principles derived from such a theory. These engineering principles allow us to specify, design, and analyze simulations of extremely large systems and implement them on massively parallel architectures. These ideas are illustrated by several socio-technical simulations being developed at Los Alamos National Laboratory