Novel Autonomous System-tactics leveraging Hypergames (NASH)

The United States Navy is committed to developing and fielding a large number of unmanned systems, including Large and Medium Unmanned Surface Vessels (USVs).Ā Unlike traditional manned platforms, autonomous platform effectiveness in accomplishing Navy missions is directly related to the sophistication of the autonomy software that controls the actions of the USVs.Ā Complex missions involving an intelligent human adversary require advanced levels of autonomy, enabling the autonomous vessels to anticipate and respond intelligently to adversary actions without intervention by a human operator.Ā A key enabler of this visionary capability is the use of game-theoretic approaches that not only optimize the actions of individual autonomous vessels, but provide the ability to reason over intelligent adversary actions, operate effectively as a team and synergistically with manned platforms as appropriate. Our proposed technology development will be applicable to a wide variety of autonomous vehicle missions and focused on robustness against an intelligent adversary. Here we propose to focus on the application of augmenting autonomous USVs with game theoretic responses towards an Anti-Submarine Warfare (ASW) mission. The proposed work will develop a game-theoretic approach to optimizing collaborative ASW tactics for teams of USVs, and will evaluate the effectiveness of those tactics against intelligent adversary models. ĀAn intelligent adversary motivates the need for a particular branch of game theory, called hyper-games, which explicitly models and reacts to the perception capabilities of both sides.Ā By adding this feature, the end system can reason over and instantiate novel strategies such as deception for a more effective set of approach options.Ā We consider the scenario of submarines passing through a monitored strait or other trip wire.Ā Our implementation will focus on the key capability of autonomous vehicles to react dynamically to the ongoing situation with very low latency. Specific collaborative behavior functionality will include: pre-detection planning laydown and collaborative search, proactive and reactive autonomous strategies, and autonomous USV behavior set to respond to detected targets.Ā These behaviors will be deployed on-board the autonomous team to enable rapid collaborative reaction and proactive responses to evolving situations.Ā In addition, the proactive/reactive behaviors will be IAW to human guidance (preferences) and constraints.