CAREER: Semantic and Goal-oriented Status Updating for Real-time Inference, Monitoring, and Decision-Making

职业：语义和目标导向的状态更新，用于实时推理、监控和决策

• 批准号: 2239677
• 负责人: Yin Sun
• 金额: $ 51.6万
• 依托单位: Auburn University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-05-01 至 2028-04-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2239677&HistoricalAwards=false
• 关键词: CAREER; Semantic; Goal; oriented; Status

Information usually has the greatest value when it is fresh. For instance, timely inference about the trajectories of nearby vehicles and pedestrians is imperative in autonomous driving, and real-time knowledge about the movements of surgical robots is essential for remote surgery. In these examples, the freshness of information is tightly related to its significance for the application. "Freshness" is an example of the "semantic" properties of the information. "Semantics" are properties that relate to the "goal" and "purpose" of information, which is to improve the performance of systems using the information. In semantic and goal-oriented status updating, the design objective is no longer sending as much as possible (i.e., increase throughput), or as fast as possible (i.e., reduce delay), without regard to the intended use of information. Instead, the transmitter should choose the right piece of information that is important for the system operating at the receiving end, and try to deliver the information to the receiver when it is most useful. This project provides innovative and foundational approaches to interpreting the freshness and semantics of information in a way that is both rigorous in theory and meaningful in practice, by characterizing the significance of such information in real-time inference, monitoring, and decision-making. Application-centric communications and scheduling techniques are developed to optimize the performance of these real-time systems. The research outcomes are expected to significantly impact numerous applications, ranging from artificial intelligence (AI) and robots to health, safety, and security, all of which are of paramount importance for the Nation's future. This project will also provide AI education and research experiences to K-12 students, undergraduate students, and students from underrepresented groups. This project develops novel semantic and goal-oriented status-updating techniques for a broad range of real-time inference, monitoring, and decision-making systems. The main goals of this project are (i) to use information-theoretic analyses and experimental studies to interpret how the freshness and semantics of information affect the performance of these real-time systems, (ii) to design new communication techniques and networking protocols for optimizing freshness/semantic measures and the system performance, and (iii) to explore the benefits of signal-aware status updating that goes beyond Age of Information to realize additional performance gain. The new status-updating techniques will be implemented to evaluate their practical performance. These research efforts are likely to lead to far-reaching impacts in the relationship of networking, machine learning, communications, and information theory at a fundamental level. 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.

信息通常在新鲜时具有最大价值。例如，在自动驾驶中，及时推断附近车辆和行人的轨迹是必不可少的，而关于手术机器人运动的实时知识对于远程手术至关重要。在这些示例中，信息的新鲜度与其对应用程序的重要性密切相关。“新鲜度”是信息的“语义”属性的一个例子。“语义”是与信息的“目标”和“目的”相关的属性，其目的是提高使用信息的系统的性能。在语义和面向目标的状态更新中，设计目标不再是尽可能多地发送（即，增加吞吐量），或者尽可能快地（即，减少延迟），而不考虑信息的预期用途。相反，发送器应该选择对接收端的系统操作重要的正确信息，并尝试在最有用的时候将信息传递给接收器。该项目提供了创新和基础的方法来解释信息的新鲜度和语义，在理论上严格，在实践中有意义，通过表征这些信息在实时推理，监控和决策中的重要性。以应用为中心的通信和调度技术的开发，以优化这些实时系统的性能。研究成果预计将对许多应用产生重大影响，从人工智能（AI）和机器人到健康，安全和安保，所有这些都对国家的未来至关重要。该项目还将为K-12学生、本科生和来自代表性不足群体的学生提供人工智能教育和研究经验。 该项目开发了新的语义和面向目标的状态更新技术，用于广泛的实时推理，监控和决策系统。该项目的主要目标是（i）使用信息理论分析和实验研究来解释信息的新鲜度和语义如何影响这些实时系统的性能，（ii）设计新的通信技术和网络协议，以优化新鲜度/语义测量和系统性能，以及（iii）探索超越信息时代的信号感知状态更新的益处，以实现额外的性能增益。将采用新的状态更新技术，以评价其实际业绩。这些研究工作可能会在基础层面上对网络、机器学习、通信和信息理论的关系产生深远的影响。 该奖项反映了NSF的法定使命，并被认为是值得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的支持。

项目成果

Learning and Communications Co-Design for Remote Inference Systems: Feature Length Selection and Transmission Scheduling

• DOI: 10.1109/jsait.2023.3322620
• 发表时间: 2023-08
• 期刊: IEEE Journal on Selected Areas in Information Theory
• 作者: Md Kamran Chowdhury Shisher;Bo Ji;I.-Hong Hou;Yin Sun

A Whittle Index Policy for the Remote Estimation of Multiple Continuous Gauss-Markov Processes over Parallel Channels

并行通道上多个连续高斯-马尔可夫过程远程估计的 Whittle 指数策略

• DOI: 10.1145/3565287.3610263
• 发表时间: 2023
• 期刊: ACM
• 作者: Ornee, Tasmeen Zaman;Sun, Yin
• 通讯作者: Sun, Yin

Age-optimal multi-flow status updating with errors: A sample-path approach

有错误的年龄最佳多流状态更新：样本路径方法

• DOI: 10.23919/jcn.2023.000041
• 发表时间: 2023
• 期刊: Journal of Communications and Networks
• 影响因子: 3.6
• 作者: Sun, Yin;Kompella, Sastry
• 通讯作者: Kompella, Sastry

Context-aware Status Updating: Wireless Scheduling for Maximizing Situational Awareness in Safety-critical Systems

• DOI: 10.1109/milcom58377.2023.10356278
• 发表时间: 2023-10
• 期刊: MILCOM 2023 - 2023 IEEE Military Communications Conference (MILCOM)
• 作者: Tasmeen Zaman Ornee;Md Kamran Chowdhury Shisher;Clement Kam;Yin Sun

A Transfer Learning-Based Deep Convolutional Neural Network for Detection of Fusarium Wilt in Banana Crops

基于迁移学习的深度卷积神经网络用于检测香蕉作物中的镰刀菌枯萎病

• DOI: 10.3390/agriengineering5040146
• 发表时间: 2023
• 期刊: AgriEngineering
• 影响因子: 2.8
• 作者: Yan, Kevin;Shisher, Md Kamran;Sun, Yin
• 通讯作者: Sun, Yin