CCSS: Learning-Driven Scheduling and Communications in Edge-Assisted Battery-Free Wireless Sensor Networks

CCSS：边缘辅助无电池无线传感器网络中的学习驱动的调度和通信

• 批准号: 2011845
• 负责人: Wei Li
• 金额: $ 38.09万
• 依托单位: Georgia State University Research Foundation, Inc.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-07-01 至 2024-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2011845&HistoricalAwards=false
• 关键词: CCSS; Learning; Driven; Scheduling; Communications

The recent significant progresses of information-sensing techniques, wireless communications, and embedded systems have greatly accelerated the generation of sensory data, which provides a nourishing fertile ground for the development of deep learning and machine learning in data-hungry applications. Nevertheless, the inherent limitations of traditional Wireless Sensor Networks (WSNs) including limited lifetime, difficult battery replacement, and centralized network architecture, become an unavoidable obstacle to the wide deployment and adoption of sensory data. Moreover, due to the big volume and high complexity of sensory data, processing sensory data in a centralized manner would increase the consumption of network resources and the risk of privacy leakage. To tackle the aforementioned limitations as well as to satisfy the needs of managing massive sensory data in real applications, developing power-optimized, sustainably-reliable, and efficiently-distributed solutions has become an essential task.This project explores the energy characteristics of battery-free WSNs, capacities of edge-assisted sinks, and advantages of distributed multi-task learning, which will result in the following technical innovations. (1) The seamless integration of battery-free sensors, edge/cloud computing and machine learning can break through technique imprisonments in traditional WSNs, in which new problems are defined and new methodologies are developed. (2) The energy correlation of battery-free sensors in the temporal and spatial domains is exploited to predict sensor’s dynamic sensing ability for scheduling sensing activities, in which the schemes of time-energy-correlated sensing, time-space cooperative data acquisition and energy-accompanied data acquisition will be developed. (3) The interference of battery-free sensors in the temporal, spatial and energy domains is utilized to construct multi-dimensional conflict graphs for interference-free transmission scheduling, in which the algorithms of associating battery-free sensors with edge-assisted sinks and scheduling data collection from sensors to sinks will be designed. (4) The diverse capacities of computation and communication on edge-assisted sinks are employed to schedule learning process so that multiple related tasks can be simultaneously completed in a distributed fashion. (5) The validation is well planned, where an analog simulator and a prototype system will be built to perform the designed simulations and real-data experiments, respectively.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

近年来信息传感技术、无线通信和嵌入式系统的重大进展极大地加速了传感数据的产生，这为深度学习和机器学习在数据饥渴应用中的发展提供了肥沃的土壤。然而，传统的无线传感器网络（WSNs）固有的局限性，包括有限的寿命，电池更换困难，集中式网络架构，成为传感器数据的广泛部署和采用不可避免的障碍。此外，由于感知数据量大且复杂度高，集中处理感知数据会增加网络资源消耗和隐私泄露风险。为了解决上述问题，并满足真实的应用中管理大量传感数据的需求，开发功耗优化、可持续可靠、高效分布式的解决方案已成为一项重要任务。本项目探讨了无电池无线传感器网络的能量特性、边缘辅助接收器的能力以及分布式多任务学习的优势，这将导致以下技术创新。(1)无电池传感器、边缘/云计算和机器学习的无缝集成可以突破传统无线传感器网络中的技术限制，定义新的问题并开发新的方法。(2)利用无电池传感器在时间和空间域的能量相关性来预测传感器的动态感知能力，从而实现时间能量相关感知、时空协同数据采集和能量伴随数据采集等感知活动的调度。(3)利用无电池传感器在时间、空间和能量域上的干扰，构建无干扰传输调度的多维冲突图，设计无电池传感器与边辅助sink的关联算法和从传感器到sink的数据收集调度算法。(4)边缘辅助汇的不同能力的计算和通信调度学习过程，使多个相关的任务可以同时完成，以分布式的方式。(5)该验证计划周密，将建立一个模拟模拟器和一个原型系统，分别执行设计的模拟和真实数据实验。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Towards Neural Network-Based Communication System: Attack and Defense

• DOI: 10.1109/tdsc.2022.3203965
• 发表时间: 2023-07
• 期刊: IEEE Transactions on Dependable and Secure Computing
• 影响因子: 7.3
• 作者: Zuobin Xiong;Zhipeng Cai;Chun-qiang Hu;Daniel Takabi;Wei Li

Privacy Threat and Defense for Federated Learning With Non-i.i.d. Data in AIoT

• DOI: 10.1109/tii.2021.3073925
• 发表时间: 2022-02-01
• 期刊: IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
• 影响因子: 12.3
• 作者: Xiong, Zuobin;Cai, Zhipeng;Li, Wei
• 通讯作者: Li, Wei

Privacy-Preserving Mechanisms for Multi-Label Image Recognition

• DOI: 10.1145/3491231
• 发表时间: 2022
• 期刊: ACM Trans. Knowl. Discov. Data
• 作者: Honghui Xu;Zhipeng Cai;Wei Li

Multi-Source Adversarial Sample Attack on Autonomous Vehicles

• DOI: 10.1109/tvt.2021.3061065
• 发表时间: 2021-03
• 期刊: IEEE Transactions on Vehicular Technology
• 影响因子: 6.8
• 作者: Zuobin Xiong;Honghui Xu;Wei Li;Zhipeng Cai

Federated Generative Model on Multi-Source Heterogeneous Data in IoT

• DOI: 10.1609/aaai.v37i9.26252
• 发表时间: 2023-06
• 作者: Zuobin Xiong;Wei Li;Zhipeng Cai