ODETTE - Automated Model Learning and Simulation of Carrier Deck Operations

Unmanned and partially autonomous aerial systems (UAS) and intelligent decision support systems have become integral components of emerging large-scale cyber-physical systems that will be involved in future military operations. Aircraft carrier decks will be central to many such operations involving multiple UAS, traditional piloted aircraft, sensory equipment, and human operators. Aurora Flight Sciences proposes to develop an algorithmic toolkit, called ODETTE (Observation-based Discrete Event Training and Test Environment), for automatically modeling carrier deck operations based on observed traffic flow data. The proposed research will enable analyses and simulations of future carrier deck operations involving new elements such as UAS and autonomous decision support systems. The proposed research is summarized as follows: We will develop a Markov Decision Process model of carrier deck operations, and survey various existing algorithms for learning of unknown state transition relations in MDPs. To address scalability, we will investigate modifications of these algorithms that are incremental and operate in a multi-resolution framework. Finally, we will develop a relatively high-fidelity software simulation of carrier deck operations, including the simulation of various sub-systems and human operators involved. This simulation environment will be used throughout the duration of the proposed research for validation and testing of the model generator.