Intelligent and distributed multi-objective methods for optimization and control of multiagents/cooperative systems

用于多智能体/协作系统优化和控制的智能分布式多目标方法

• 批准号: RGPIN-2021-03737
• 负责人: Blondin, MaudeJosée
• 金额: $ 2.4万
• 依托单位: Université de Sherbrooke
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2021
• 资助国家: 加拿大
• 起止时间: 2021-01-01 至 2022-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=742698
• 关键词: Intelligent; distributed; multi; objective; methods

Large teams of robots can accomplish complex tasks such as land-mine clearance. To do so, researchers from robotics to computer science have developed algorithms to have each agent in the multiagent robotic swarm make smart choices in optimizing several objectives. At this level, multi-objective optimization techniques will make multiagent systems more efficient and customized to accomplish the tasks at hand, saving time, life, and energy; however, the existing theory is falling behind. Indeed, in multiagent problems with many objectives, the existing literature considers mainly problems in which agents optimize the diverse functions with equal priorities; however, there exist many cases in which there are objectives with different importance. For example, teams of robots may want to explore different regions of an area, or agents may have different priorities in trajectory planning when minimizing both energy consumption and travel time. Therefore, the next breakthrough in multi-objective optimization multiagent problems is to capacitate agents to prioritize objectives individually.  Thus, through distributed multi-objective optimization, the research program's objective is enhancing the decision-making skills of multiagent systems to enable new practical features, increase agent capacities, and provide a broader range of operating conditions of such systems. To support the research program's objective and help with the next breakthrough, this discovery grant consists of four innovative and independent projects but that are interconnected. In particular, the first five years of this research program will focus on developing distributed multi-objective optimization methods, addressing the existing theoretical limitations in a multiagent context, including in swarm intelligence techniques, and integrating distributed multi-objective optimization methods into foraging and target search tasks. The proposed program will have a significant impact on various fields. From a technological and humanitarian standpoint, with the developed algorithms, land-mine clearance robots will independently prioritize areas requiring more exploration according to external data received in real-time by the agents. It will boost exploration efficiency, resulting in saving more lives. Moreover, the developed algorithms will improve foraging tasks performed by multi-robot systems used in many applications such as surface chemical skimming of unintentional oil spills. From an environmental and social standpoint, the algorithms will dispatch the distributed energy resources in smart-grid more efficiently, resulting in saving money for the consumers and the stakeholders and preserving the environment. Also, in the domain of intelligent transportation systems, the algorithms will be useful tools for distributed routes planning.

大型机器人团队可以完成复杂的任务，如扫雷。为了做到这一点，从机器人学到计算机科学的研究人员开发了算法，让多智能体机器人群体中的每个智能体在优化几个目标时做出明智的选择。在这个层面上，多目标优化技术将使多智能体系统更加高效和定制，以完成手头的任务，节省时间、生命和能源；然而，现有的理论正在落后。事实上，在具有多个目标的多智能体问题中，现有文献主要考虑了智能体以相等的优先级优化不同功能的问题；然而，在许多情况下，存在不同重要性的目标。例如，机器人团队可能想要探索一个区域的不同区域，或者代理在最小化能源消耗和旅行时间时，在轨迹规划方面可能有不同的优先级。因此，多智能体优化问题的下一个突破口是使智能体能够单独地对目标进行优先排序。因此，通过分布式多目标优化，研究计划的目标是提高多智能体系统的决策技能，以使其具有新的实用特征，增加智能体的能力，并为此类系统提供更广泛的运行条件。为了支持研究计划的目标，并帮助实现下一步的突破，这项发现拨款由四个相互关联的创新和独立项目组成。特别是，这个研究计划的前五年将专注于开发分布式多目标优化方法，解决多智能体环境中现有的理论限制，包括在群体智能技术中，并将分布式多目标优化方法集成到觅食和目标搜索任务中。拟议中的计划将对各个领域产生重大影响。从技术和人道主义的角度来看，有了开发的算法，扫雷机器人将根据代理人实时收到的外部数据，独立地对需要更多勘探的地区进行优先排序。它将提高勘探效率，从而拯救更多的生命。此外，开发的算法将改善多机器人系统执行的觅食任务，这些系统用于许多应用，如意外漏油的表面化学撇油。从环境和社会的角度来看，该算法将更有效地调度智能电网中的分布式能源资源，从而为消费者和利益相关者节省资金，保护环境。此外，在智能交通系统领域，这些算法将成为分布式路线规划的有用工具。