An agent based compositional framework for supply chain simulation

To survive in an ever increasing global and competitive marketplace, organisations are forging strategic alliances to gain a competitive advantage over their rivals. Consequently, it is now recognised that it is not sufficient to look at organisations in isolation, but view them in the wider context of the supply chain. In order to design arid manage supply chains it is necessary to understand and predict the behaviour of such systems. The ability to perform detailed studies of dynamic behaviour has made discrete event simulation (DES) an invaluable tool in the design and analysis of manufacturing systems. DES has been used to model individual stages of a supply chain, but rarely has it been applied comprehensively across the entire chain. The multi-faceted nature of supply chains makes the creation of a single model that represents all aspects of the chain difficult. A compositional framework, termed HerMIS (Heterogeneous Model Integration and Simulation), is proposed that allows pieces of a supply chain to not only be studied in isolation, but in the context of the other parts as well. Three requirements are identified for the development of HerMIS. These are: (1) to support a compositional approach so as to allow multi-facetted modelling, (2) to function in a distributed environment where models and information about them are distributed at different locations amongst various organisations, and (3) to provide an execution mechanism that allows the composite model to be simulated efficiently. A class based taxonomy of component models and their interaction is conceived that forms the basis of a representation scheme for composite modelling. An agent based paradigm that employs a collection of synthesis_agents and model_agents is devised to support the distributed operation of the framework. The synthesis_agents function as sources of knowledge for synthesising composite models and are used in conjunction with an interactive blackboard based system to guide the user in creating composite models. Each of the model_agents incorporate a discrete event model of a supply chain component, arid supports the distributed simulation of the composite model. Finally, a parallel discrete event simulation algorithm is proposed that enables the composite model to be simulated on a network of computer workstations. The algorithm is based on the optimistic PDES approach and takes into consideration some of the operating characteristics of a composite supply chain model.