Architecture to Support Interactive Training Simulation

Intelligent Automation, Inc. (IAI) and our collaborators propose to apply IAI’s CybeleTM agent-based system to develop an architecture that supports interactive, training simulations on supercomputers. The CybeleTM agent based system already incorporates many of the characteristics required for an effective middleware for executing simulations on supercomputers: CybeleTM is inherently distributed and may be hosted on each of the nodes in the supercomputer cluster; it supports threads of parallel activity, and it interacts between components (agents) by passing messages. To develop this architecture, we will (1) work with UCF/IST supercomputer scientists to understand characteristics of their supercomputer; (2) work with designers and developers of simulation based training system to understand their requirements and limitations when using standard computers; (3) design an architecture that meets the requirements of interactive training simulations; and (4) begin to implement the architecture for a specific training system. The option and Phase II would complete implementations of the architecture for interactive training systems. We foresee inviting interested organizations with interesting training simulations to implement their systems on the supercomputer. This project should result in implemented middleware for running interacting training simulations on supercomputers, and guidance for building or modifying training simulations to make effective use of supercomputers.