| The thesis of this dissertation is that execution of the actions required to accomplish a task specified by an agent program is similar to an operating system executing a sequence of user requests. In the same fashion that an operating system provides an environment for the execution of a user request, an agent framework provides the needed environment for the execution of agent actions. The agent environment includes the ability to communicate with other agents, maintaining the current state of an executing agent, and selecting an execution path from a set of possible execution paths. The specific contributions of this work are: a framework for the execution of intelligent software agents; a characterization of agent performance; development of a real-time agent scheduling algorithm; development of an environment for testing and comparing the performance of agent activities, and a library of benchmarks for testing and comparing scheduling algorithms and architectures.; The agent architecture is called DECAF (Distributed Environment Centered Agent Framework). DECAF is a software toolkit for the rapid design, development, and execution of “intelligent” agents to achieve solutions in complex software systems. Unlike a traditional operating system, an agent architecture has the ability to reason about action execution if a characterization of action performance is available. A major component of the described work is a characterization of agent execution that will lead to optimal or near optimal scheduling of agent execution. This document will describe the key portions of the DECAF toolkit, the internal details of the agent execution framework, and a formal model of the parameters used for scheduling considerations.; The scheduling algorithm is called DRU (Dynamic Real-time Updates). DRU features the ability to examine schedules and execute several task structures in parallel that will accomplish a goal. Additionally, DRU monitors the progress of actions within tasks and assures that deadlines are met even in the event of specific failures. Additionally, automatic generation of a suite of test procedures is described along with actual multi agent systems developed using DECAF and DRU. |
