Abstract
While the US Army is adopting a Modular Open Systems Approach (MOSA) to building its future Enterprise Modeling and Simulation (M&S), reaching that future is following an evolutionary approach that requires us to smartly leverage existing simulation systems that are often diverse and complex. The use of simulation varies in scenario, available simulation services, simulation architecture (e.g. middleware used), computing power available, and network infrastructure, among other considerations. As increasingly complex simulation systems are composed, deployment of these systems also increases in complexity and presents many challenges. The move towards a cloud-compliant solution eases the burden to some extent; however, simulation deployments still require careful planning by a team of experts, which typically requires effort for composition and deployment for each simulation exercise.
In this paper, we present our approach to optimize and automate the deployment of distributed simulation systems that we have researched and developed as part of the Automated Simulation Management (ASM) project. We will detail our optimizer that takes into consideration different simulation architectures (i.e. single location, multiple sites, Modeling and Simulation as a Service (MSaaS)) and deployment platforms (i.e. local machines, cloud, and hybrid) to provide a solution for before and during run-time. Through a machine understanding of the scenario, service composition, (exercise) use case, and target environment, we automate the deployment of the simulation environment. The core of the solution is a genetic algorithm that processes data from a knowledge model and provides optimal deployment results. Finally, these results are processed by a Large Language Model (LLM)-based script generator to create an environment specific configuration. The result is an effective orchestration of simulation capabilities in order to meet the simulated goals balancing models/simulations, computing, and networking.
