Mechanical Engineering Department, University of Washington, Seattle, WA 98195.
Proc Natl Acad Sci U S A. 2023 Mar 14;120(11):e2219948120. doi: 10.1073/pnas.2219948120. Epub 2023 Mar 10.
A method for low-distortion (low-dissipation, low-dispersion) information propagation in swarm-type networks with suppression of high-frequency noise is presented. Information propagation in current neighbor-based networks, where each agent seeks to achieve a consensus with its neighbors, is diffusion-like, dissipative, and dispersive and does not reflect the wave-like (superfluidic) behavior seen in nature. However, pure wave-like neighbor-based networks have two challenges: i) It requires additional communication for sharing information about time derivatives and ii) it can lead to information decoherence through noise at high frequencies. The main contribution of this work is to show that delayed self-reinforcement (DSR) by the agents using prior information (e.g., using short-term memory) can lead to the wave-like information propagation at low-frequencies as seen in nature without the need for additional information sharing between the agents. Moreover, it is shown that the DSR can be designed to enable suppression of high-frequency noise transmission while limiting the dissipation and dispersion of (lower-frequency) information content leading to similar (cohesive) behavior of agents. In addition to explaining noise-suppressed wave-like information transfer in natural systems, the result impacts the design of noise-suppressing cohesive algorithms for engineered networks.
提出了一种在具有高频噪声抑制的群集型网络中实现低失真（低耗散、低色散）信息传播的方法。当前基于邻居的网络中的信息传播类似于扩散，具有耗散性和色散性，并且不能反映自然界中看到的波状（超流）行为。然而，纯波状基于邻居的网络有两个挑战：i）它需要额外的通信来共享关于时间导数的信息；ii）它可能会通过高频噪声导致信息去相干。这项工作的主要贡献是表明，通过使用先前的信息（例如，使用短期记忆）进行代理的延迟自增强（DSR）可以在没有代理之间的额外信息共享的情况下导致类似自然界中看到的低频的波状信息传播。此外，结果表明，DSR 可以设计为抑制高频噪声传输，同时限制（较低频率）信息内容的耗散和色散，从而导致代理的类似（凝聚）行为。除了解释自然系统中抑制噪声的波状信息传输外，该结果还影响了用于工程网络的抑制噪声的凝聚算法的设计。
