Development of a Distributed Simulation Cluster for MOOS-IvP

Abstract

Batch testing simulations of autonomy software make verification and optimization much easier and more robust. The ability to verify and optimize code is particularly important for large, expensive assets, such as marine vessels, where the cost of any failure is high. This thesis discusses the architecture, implementation, and use of an expandable simulation toolbox for MOOS-IvP. The toolbox utilizes Monte Carlo simulations since these are incredibly flexible for differing scenarios. A distributed architecture improves robustness since a single failure does not bring the cluster down. Additionally, personal computers may be added to the cluster during off hours, thus increasing the average computing power.