RI: Small: RUI: Fission-Fusion Multi-Robot Systems

在探索、侦察、搜索和救援等现实任务中使用机器人团队的能力取决于它们在复杂和动态环境中有效合作的能力。不幸的是，为给定任务确定最有效的团队规模取决于各种因素，通常需要不易获得的信息，并且在计算上可能不切实际。来自自然界的观察表明，一些动物社会表现出频繁的分裂（裂变）和融合（融合）的子群体，而没有多机器人系统中常见的协调开销。本研究项目的目标是利用自然界裂变融合社会的这些见解作为灵感，在多机器人系统中实现类似的行为。这个项目将调查动机和机制，有助于一个团队的机器人裂变成更小的群体和融合成更大的群体，为适当的规模给定的任务。潜在的生物和环境因素有助于个体决策，导致群体的裂变和融合，并具有与多机器人系统相关的人工类似物，将使用基于代理的模拟来识别。重点将放在有助于适应特定任务的首选群体规模的因素上。一旦确定，这些因素将在物理机器人上实施，以评估物理机器人约束下决策系统的性能。所使用的决策系统将使用自适应模糊行为层次和利用自然系统观察开发的集体决策模型来实现。

The ability to use teams of robots in real-world tasks, such as exploration, reconnaissance and search and rescue, depends on their ability to effectively cooperate in complex and dynamic environments. Unfortunately, determining the most effective team size for a given task depends on a variety of factors, often requires information not readily available, and can be computationally impractical. Observations from nature show that some animal societies exhibit frequent splits (fission) and merges (fusion) of subgroups without the coordination overhead frequently found in multi-robot systems. The goal of this research project is to use these insights from fission-fusion societies in nature as inspiration to implement similar behavior in multi-robot systems.This project will investigate the motivations and mechanisms that contribute to the fission of a team of robots into smaller groups and the fusion of smaller groups into a larger group, into sizes appropriate to a given task. Potential biological and environmental factors that contribute to individual decisions that result in the fission and fusion of groups and that have artificial analogues relevant to multi-robot systems will be identified using agent-based simulations. An emphasis will be placed on factors that contribute to an adaptive task-specific preferred group size. Once identified, these factors will be implemented on physical robots to evaluate the performance of the decision-making system within the constraints of physical robots. The decision-making system used will be implemented using an adaptive fuzzy behavior hierarchy and a collective decision-making model developed using observations of natural systems.