From Compressed Sensing to Collective Sensing: a Complex Network Approach

李欣，西弗吉尼亚大学从压缩感知到集体感知：一种复杂的网络方法智力优势：基于还原论的分而治之方法在从传感器和手机到机器人和计算机的许多工程系统的设计中都取得了丰硕的成果。相比之下，自然界中的复杂系统，如蚁群和人脑，是由大量简单单位相互作用以及环境的集体动力学驱动的。这些单位如何以自组织的方式形成全球复杂的行为，仍然是科学界的一个大谜团。解决这一难题，甚至部分解决，可能会对集成和混合系统的设计和优化产生深远的影响，从而更好地服务于各种工程应用。这个项目的重点是对传感和处理集成的集体方法。与压缩感知不同，压缩感知的成功是基于数学规范或基函数的巧妙构建，该项目利用了物理世界中感官信号的潜在组织原则。在这个项目的成功完成后，将有一个更好的理解复杂网络的集体动态如何与传感组件相互作用，以支持复杂的任务，如运动感知和模式识别。更广泛的影响：建议的研究活动将通过培养研究生（包括科学研究的伦理），促进网络科学相关的学习，开发网络空间教材，以及扩大代表性不足群体的参与，融入教育。所有收集的数据、生成的代码和实验结果都将向公众发布，这促进了可重复性研究的原则。对工程的科学观点有望使工程专业的学生在课堂内外受益。网络科学与传感器技术之间的联系可以利用到电子成像行业，并对集成传感器系统的设计产生潜在的影响。

ECCS-0968730Xin Li, West Virginia UniversityFrom Compressed Sensing to Collective Sensing: a Complex Network ApproachABSTRACTIntellectual Merit: Reductionism-based divide-and-conquer approach has been fruitful to the design of many engineering systems from sensors and cell-phones to robots and computers. By contrast, complex systems in nature such as ant colony and human brain are driven by the collective dynamics involving a large number of simple units interacting with each other as well as the environment. How those units form a globally complex behavior in a self-organizing fashion remains a big mystery in science. Solving this puzzle, even partially, may have profound implications into the design and optimization of integrative and hybrid systems that could better serve various engineering applications. This project focuses on a collective approach toward the integration of sensing and processing. Unlike compressed sensing whose success is based on ingenious construction of mathematical norms or basis functions, this project exploits the underlying organizational principle of sensory signals in the physical world. Upon the successful completion of this project, there will be an improved understanding of how the collective dynamics of complex networks interacts with the sensing component to support complex tasks such as motion perception and pattern recognition.Broader Impacts: The proposed research activities will be integrated into the education through training graduate students (including ethnics of scientific research), promoting network-science related learning, developing cyber-space teaching material, and broadening the participation of underrepresented groups. All collected data, produced codes and experimental results will be released to the public, which promotes the principle of reproducible research. The scientific perspective toward engineering is expected to benefit engineering students both inside and outside the classroom. The connection between network science and sensor technology could leverage into the industry of electronic imaging and have potential impact on the design of integrated sensor systems.