CIF: SMALL: Explorations and Insights into Adaptive Networks, Animal Flocking Behavior, and Swarm Intelligence

自20世纪90年代初以来，社会科学和生物科学在研究动物群体行为和群体智能方面出现了一些有用的优化算法。例如，已经观察到，虽然动物群体中的单个代理人不能做出复杂的行为，但多个代理人之间的联合协调导致了行为规律的表现。已经开发了几种算法来模拟动物群体的运动。事实证明，这些研究在建模和理解复杂现象以及开发从生物学到纳米技术等领域的应用方面都很有用。这项研究涉及调查生物和社会科学中关于群体智能的这些研究与最近在系统论中关于适应网络的研究之间的相互联系。自适应网络由分布在地理域上的各向同性节点组成。节点感知环境，并尝试基于它们的噪声观测来理解感兴趣的现象，而不需要任何节点扮演中央控制角色。节点通过本地交互相互协作，并根据在节点处收集的数据和从邻居接收的数据来调整它们的状态和网络拓扑。到达节点的信息通过扩散过程在整个网络中传播。信息的扩散导致了一种集体智能的形式，相对于非合作网络，改进的学习和收敛行为证明了这一点。对群体行为和自适应网络的动力学进行更深入的研究，可以为设计自适应网络和理解群体行为提供替代技术，并可能对这些发展所激发的应用产生潜在影响。

Abstract Since the early 1990s, some useful optimization algorithms have emerged from the social and biological sciences in their studies of animal flock behavior and swarm intelligence. It has been observed, for example, that while individual agents in an animal colony are not capable of complex behavior, the combined coordination among multiple agents leads to the manifestation of regular patterns of behavior. Several algorithms have been developed to model the movement of animal flocks. These investigations are proving useful in modeling and understanding complex phenomena and in developing applications in areas ranging from biology to nanotechnology. The research involves investigating interconnections between these studies on swarm intelligence in the biological and social sciences, and more recent studies on adaptive networks in system theory. Adaptive networks consist of isotropic nodes spread over a geographic domain. The nodes sense the environment and attempt to understand a phenomenon of interest based on their noisy observations and without any node taking a central control role. The nodes cooperate with each other through local interactions and adapt their states, and the network topology, in response to data collected at the nodes and data received from their neighbors. Information arriving at a node propagates throughout the network by means of a diffusive process. The diffusion of information results in a form of collective intelligence as is evidenced by improved learning and convergence behavior relative to non-cooperative networks. A closer study of the dynamics of swarm behavior and adaptive networks can suggest alternative techniques for designing adaptive networks and for understanding flock behavior, with potential impact on the applications that can be motivated from these developments.