Enterprise Dynamics Via Non-Equilibrium Membrane Models Wolf Kohn Hynomics Corp., 10632 NE 37th Circle, Building 23, Kirkland, WA 98033-7021 & wk@hynomics.com Vladimir Brayman Department of Electrical Engineering, University of Washington, Campus Box 352500, Seattle, WA 98195-2500 & vbrayman@ee.washington.edu James A. Ritcey Department of Electrical Engineering, University of Washington, Campus Box 352500, Seattle, WA 98195-2500 & ritcey@ee.washington.edu (Received 2000) Abstract. This paper describes a distributed dynamic model of enterprise systems via a network of elements which are abstractions of biological membranes. Membrane characteristics such as active sites controlling the flow of substances correspond to local feedback laws in the elements of the supply chain of the enterprise. Flow conservation and chemical reactions of substances across the membrane are abstracted to represent component flow interaction in the supply chain. The model characteristics are illustrated with a simulation example. This model methodology is completely encodable. It provides a blueprint for highly automated model generation of enterprise systems, and for on-line generation of continuous repair implementations of planning, scheduling and execution applications. The proposed embedded distributed control system allows for the realization of diverse optimization strategies because a given criterion is approximated by a generic criterion via the penalty method. The control system also satisfies network element constraints and inter-element synchronization requirements. 1. Introduction In November of 1993, when we first considered the application of Networks of Hybrid Control agents to economics problems [1], we were at the stage of ideas. In that work we modeled the economic activity of a firm as an entity which is simultaneously playing several different types of classical economic games, referred to as base games. The selected base games were min-max, Pareto, Stackelburg, Team, and Nash. We showed that the hybrid systems theory, applied to this game situation, coupled with the underlying mathematical theory of our hybrid agents [2], allows the generation of an optimization policy that corresponds to the chattering combination of the policies optimizing the base games. We discovered that this modeling approach was deficient in several ways. Par- Exemplary OSID style