BIGDATA: F: Collaborative Research: Acquisition, Collection and Computation of Dynamic Big Sensory Data in Smart Cities

信息传感设备的无处不在为大传感数据（BSD）开辟了丰富的来源，其跨越物联网，无线传感器网络，RFID，网络物理系统等。这种多样化的bsd丰富的系统是智能城市的基石，智能设备部署在城市的每个角落。BSD的分析使用对于智慧城市管理城市资产和监测空气状况、污染、气候变化、交通、安全等至关重要。对智慧城市的高需求和传感设备在智慧城市中的关键作用加速了BSD的爆炸式增长。不幸的是，BSD的大小和动态特性压倒了当前捕获、存储、搜索、挖掘和可视化BSD的能力，因此已成为智慧城市应用程序广泛发展的主要障碍。为了应对这些挑战，该项目将调查有关BSD获取、收集和计算的基本问题，并进行原则性的质量控制。目标是经济有效地收集和管理BSD，以便在智慧城市应用程序中有效利用。针对BSD的大规模、相关动态、模式多样性和低质量这四个具有挑战性的特点，将开发一套BSD管理的基本原则、算法和工具。这项研究的结果将有助于实现智慧和弹性城市的愿景，这将广泛影响国家新兴的智慧城市基础设施和公民的移动生活质量。该项目还提供了一个与哥伦比亚特区政府合作开展“智慧华盛顿倡议”的机会，从而不仅影响研究界，也影响整个社会。该项目旨在解决在数据采集、收集和计算的关键阶段任务认知BSD管理的主要挑战。总体目标是缓解BSD在智慧城市应用中的高计算成本和提高利用效率。首先，将开发基于物理世界变化趋势自动调整传感频率的近似BSD采集方法。这样的采集方式可以在智慧城市的周期性和长期监测中，有效减少早期的数据量。其次，将为特定任务的多模态BSD开发近似采样算法、知识发现方法和集成方法，以减少从传感设备向最终用户提供原始和冗余BSD相关的传输成本。最后，将研究评估BSD质量的新指标，然后将其应用于正确评估低质量BSD的容忍度，并深入了解数据质量对BSD获取、收集和计算的各个设计方面的基本影响。除了理论分析外，还将在实际的BSD上进行模拟和实验研究，包括在华盛顿特区的现实智能城市项目上进行实验。相应的代码、数据集和教育材料将通过一个专门的项目网站发布。

The ubiquity of information-sensing devices has opened up abundant sources for Big Sensory Data (BSD), which span over Internet of Things, wireless sensor networks, RFID, cyber physical systems, to name a few. Such diverse BSD-rich systems are the building blocks for smart cities where smart devices are deployed in every corner of a city. The analytical use of BSD is essential to smart cities in managing a city's assets and monitoring air conditions, pollution, climate change, traffic, security and safety, etc. The high demand for smart cities and the pivotal role of sensing devices in smart cities accelerate the explosion of BSD. Unfortunately, the size and dynamic nature of BSD overwhelm current capability to capture, store, search, mine and visualize BSD, and hence have become a major hindrance to the widespread development of smart city applications. To tackle these challenges, this project will investigate fundamental issues regarding acquisition, collection and computation of BSD with principled quality control. The goal is to cost-effectively collect and manage BSD for efficient utilization in smart city applications. A set of foundational principles, algorithms and tools for BSD management will be developed in response to the four challenging characteristics of BSD, which are large scale, correlated dynamics, mode diversity and low quality. The outcomes of this research will contribute to the vision of smart and resilient cities, which broadly impact the nation's emerging smart city infrastructure and citizens' mobile quality of life. This project also offers an opportunity to collaborate with the Government of the District of Columbia on the Smarter DC Initiative, and hence impacts not only the research community but also the society at large. This project aims at tackling major challenges in task-cognizant BSD management at the critical phase of data acquisition, collection and computation. The overarching goal is to alleviate the high computational cost and improve the utilization efficiency of BSD in smart city applications. First, approximate BSD acquisition methods will be developed that automatically adjust the sensing frequency based on the changing trend of the physical world. Such acquisition methods can effectively reduce the data volume at an early stage during periodic and long-term monitoring of smart cities. Second, approximate sampling algorithms, knowledge discovery methods, and integration methods will be developed for task-specific multimodal BSD, in order to reduce the transmission cost associated with delivering otherwise raw and redundant BSD from sensing devices to end users. Finally, new metrics for evaluating BSD quality will be investigated and then applied to properly assess the tolerance of low-quality BSD and provide deep understanding of the fundamental impact of data quality on various design aspects of BSD acquisition, collection and computation. Besides theoretical analysis, simulation and experimental studies will be carried out on real BSD, including experimentation on real-world Smart City projects at Washington DC. The corresponding code, datasets, and educational materials will be released via a dedicated project website.