Generation of architectures for distributed intelligence systems

An approach to the modeling of discrete event distributed intelligence systems is presented that uses ordinary Petri Nets for fixed structure architectures. The model leads to the precise formulation of the problem of generating the complete set of architectures that satisfy a number of resource and design constraints. Two algorithms are presented: the first one (the DFS algorithm) starts with the specification of the desired functionality and generates data flow structures with prescribed redundancy and complexity; then the various functions are allocated to resources; the second one (the Lattice algorithm) starts with a given set of decision making units and obtains the partially ordered sets that contain all the feasible architectures. The resulting architectures can be analyzed to evaluate their performance characteristics.