Online Dynamic Graphical Game Approach for Pursuing Illegal Fishing Vessels with Running Energy Optimization**

通过运行能量优化追捕非法渔船的在线动态图形游戏方法**

• 批准号: 537568-2018
• 负责人: Gueaieb, Wail
• 金额: $ 1.82万
• 依托单位: University of Ottawa
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=652591
• 关键词: Online; Dynamic; Graphical; Game; Approach

"Illegal, unreported, and unregulated" (IUU) fishing events are having severe environmental and economic consequences on Canadian water resources. They lead to harmful changes in the ecosystem, depleting the natural resources, and causing major financial losses. This urges for the development of innovative solution platforms to autonomously detect, track, and pursue IUU fishing events during the monitoring of maritime territories. This leads to an interesting structured engineering problem with several technical challenges. First, enhancing the ability of each pursuing vessel to track and detect IUU fishing events in an efficient manner. Second, enhancing the capability of the pursuing vessels to communicate among each other through an effective communication or information layer that helps achieving their cooperative objectives (e.g., getting closer to the illegal vessel, while exchanging the necessary information with the neighboring pursuing vessels) and global objectives (e.g., collectively achieving the final objective by the participating vessels). Third, optimizing the performance of each pursuing vessel, by minimizing its dissipated energy during the monitoring process, for instance. These objectives can be tackled using recent advances in artificial intelligence (AI). To this end, the proposed research aims at devising a solution that brings together ideas from machine learning (ML), game theory, computational graph theory, optimal control, and neurofuzzy systems. The proposed approach is based on designing online game-theoretic approaches which take advantage of state-of-the-art reinforcement learning schemes to learn the behavior of illegal vessels. Game theory is applied to understand and model the interactive nature of the multi-agent system (the pursuing against the illegal vessels). Optimal control theory formulates the optimization problem along with the different applicable constraints leading to the optimality conditions at which the different sub-systems act in an optimized fashion. This research will help advancing the state of knowledge of ML, which is in line with the nation's strategic plan to build reputed AI hubs.

“非法、未报告和无管制”的捕捞活动对加拿大的水资源造成了严重的环境和经济后果。它们导致生态系统的有害变化，耗尽自然资源，并造成重大经济损失。这就要求开发创新的解决方案平台，以便在监测海洋领土期间自动检测、跟踪和追踪IUU捕捞事件。这导致了一个有趣的结构化工程问题，有几个技术挑战。第一，提高每艘追捕船只有效跟踪和发现非法、未报告和无管制的捕捞活动的能力。第二，通过有效的通信或信息层，提高追捕船只相互通信的能力，以帮助实现其合作目标（例如，更接近非法船只，同时与邻近的追捕船只交换必要的信息）和全局目标（例如，最终目标是通过参与的船舶共同实现）。第三，优化每艘追击船的性能，例如，在监测过程中最大限度地减少其消耗的能量。这些目标可以利用人工智能（AI）的最新进展来解决。为此，拟议的研究旨在设计一种解决方案，将机器学习（ML），博弈论，计算图论，最优控制和神经模糊系统的思想结合在一起。所提出的方法是基于设计在线游戏理论的方法，利用最先进的强化学习计划来学习非法船只的行为。博弈论被应用到理解和建模的多代理系统的互动性质（对非法船只的追求）。最优控制理论将优化问题沿着与导致不同子系统以优化方式作用的最优性条件的不同适用约束一起公式化。这项研究将有助于推进机器学习的知识状态，这符合国家建立知名人工智能中心的战略计划。