CyberOrgs: A Model for Resource Bounded Complex Agents

The ubiquity of networked computing devices combined with recent advances in computer networks technology have created the possibility of harnessing the collective computational power of peer computational resources to carry out very large computations. Agents offer a natural framework for abstracting distributed computations. However, the framework needs to be adapted to explicitly model resources and their ownership in a way that transforms a network of peer-owned resources into a distributed execution environment that can hosts peer-initiated multi-agent computations. Additionally, mechanisms are required for supporting trade in peer-owned resources as well as control of available resources. This thesis proposes CyberOrgs, as a model for resource bounded multi-agent computations over a network of peer-owned resources. CyberOrgs compose Agents with a model for resource acquisition and control, so that computations carried out by ensembles of agents along with resources committed to their execution are encapsulated inside resource boundaries. Specifically, each cyberorg owns resources and eCash to purchase additional resources with; it manages a concurrent computation being pursued by an ensemble of agents, and supports their execution by purchasing resources for them from other cyberorgs; and, it provides resources to other cyberorgs in exchange for eCash in accordance with contracts negotiated with them. A transition system has been developed to concretize the operational semantics of CyberOrgs, and some properties of the model have been examined. A prototype implementation has been developed as an Actor program which offers programming constructs to support acquisition and control of resources, while allowing a programmer to keep resource and functional concerns of an application separate. The expressive power of these constructs is illustrated using examples