CAREER: Integrated and end-to-end machine learning pipeline for edge-enabled IoT systems: a resource-aware and QoS-aware perspective

在未来的智慧城市中，人工智能和边缘计算的融合带来了大量的应用，为可持续的城市生活创造了变革性的潜力。从智能医疗系统到智能交通控制系统，这些应用程序都与地理分布设备生成的大量数据集的处理相关联。该项目的目标是开发一个集成的、可靠的管道，该管道将有效、自动地准备、清理和分析相关的分布式数据集，同时最大限度地降低系统的总体成本，并在数据准备和数据处理任务之间实现动态平衡。这将使用创新技术来完成，包括联邦数据预处理、联邦学习、新的编码方案和压缩技术。将开发一套优化问题和相关的算法解决方案。所提出的方法将通过使用PI实验室开发的测试平台进行广泛的模拟和实验来验证和改进。通过启用数据驱动的智能技术，例如智能医疗监控和智能交通控制系统，该项目有可能显著改善美国公民的生活质量，而这些技术目前尚不可行。一个关键目标是为创造更有效和可持续的城市环境作出贡献。此外，该项目将包括通过肯塔基州的一切都是科学节等当地活动进行综合教育、推广和指导活动，并与肯塔基州-西弗吉尼亚州路易斯斯托克斯少数民族参与联盟合作。一个关键目标是促进多样性和包容性，并赋予在机器学习、数据科学和边缘计算等新兴领域工作的下一代专家权力。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

In the landscape of future smart cities, the integration of artificial intelligence and edge computing has led to a multitude of applications that create transformative potential for sustainable urban living. From smart healthcare systems to intelligent traffic control systems, these applications are linked to processing of substantial datasets generated by geographically distributed devices. The objective of this project is to develop an integrated and reliable pipeline that will effectively and automatically prepare, clean, and analyze the associated distributed datasets while minimizing overall costs of the system and dynamically balancing between data preparation and data processing tasks. This will be accomplished using innovative technologies, including federated data pre-processing, federated learning, new coding schemes, and compression techniques. A suite of optimization problems and associated algorithmic solutions will be developed. The proposed methodologies will be validated and refined through extensive simulation and experiments performed using a testbed developed within the PI’s lab. This project has the potential to significantly improve the quality of life for US citizens by enabling data-driven, smart technologies, such as smart healthcare monitoring and smart traffic control systems that are not yet feasible. A key goal is to contribute to creating more efficient and sustainable urban environments. Further, the project will include integrated education, outreach, and mentoring activities through local events like the Everything is Science Festival in Kentucky, and collaborating with the Kentucky-West Virginia Louis Stokes Alliance for Minority Participation. A key goal is to foster diversity and inclusion and empower the next generation of experts working in the emerging fields of machine learning, data science, and edge computing.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.