Embedded Databases for Agricultural and Environmental Applications

环境、农业和工业应用的传感器设备收集的数据的快速增长正在挑战现有的数据处理方法。虽然数据收集和分析提高了效率和可持续性，但也存在相关成本，包括能源使用、电子和电池浪费以及部署和维护时间。我的长期研究目标是提高低功耗嵌入式传感器设备上数据库软件的效率，从而减少能源使用，降低维护成本，并延长设备使用寿命。我的研究小组已经开发了用于SQL和键值嵌入式数据库的软件，以及用于数据索引、排序和存储的高效技术。该软件运行在只有4 KB内存的设备上，并为没有标准数据库库的嵌入式应用程序提供独特的支持。在过去的五年中，参与的学生群体多样化，包括四名现在担任大学教授的毕业生，四名NSERC USRA获得者，以及二十多名在工业中担任软件工程角色的本科生。未来五年的研究目标是使数据管理算法适应最新的硬件和网络技术，并创造技术，为环境和农业用户简化传感器系统的部署。数据库算法需要对内存有限的设备进行优化，对于具有独特存储特性的非易失性存储设备（如NOR闪存），需要更有效的算法。最新的低能耗网络协议，如LoRa，允许创新的数据管理技术，因为传感器上的本地数据处理可以取代网络上的数据传输。我们将在之前开发的软件的基础上进行新的研究，通过利用数据压缩、自适应采样和机器学习技术来减少无线传输的数据量。这些技术将通过自适应地确定存储、传输和丢弃哪些数据以获得最佳性能来提高传感器效率。该项目将为理解有限资源下的数据处理做出基础性贡献，发布开源软件以提高易用性，并培养具有嵌入式软件工程技能的5名研究生和10名本科生HQP。一个主要用例是支持农业应用中收集的时间序列数据。随着气候变化和水资源短缺的影响越来越大，智能利用技术来提高可持续性，减少水和肥料的使用，增加收成至关重要。一个特别的重点是继续和扩大与基洛纳市的合作，奥肯纳根的葡萄园和农业生产者，以及环境机构，部署我们的研究成果，以解决水管理，气候监测和农业生产力的关键挑战。

The rapid growth of data collected by sensor devices for environmental, agricultural, and industrial applications is challenging existing approaches to data processing. Although data collection and analysis provide improved efficiency and sustainability, there are also associated costs including energy use, electronic and battery waste, and deployment and maintenance time. My long-term research goal is to improve the efficiency of database software on low-power embedded sensor devices resulting in reduced energy use, lower maintenance costs, and increased device lifetime. My research group has produced software for SQL and key-value embedded databases, and efficient techniques for data indexing, sorting, and storage. This software runs on devices with as little as 4 KB of memory and provides unique support for embedded applications where there is no standard database library. The diverse group of students involved in the last five years includes four graduates now working as college professors, four NSERC USRA recipients, and over twenty undergraduates with software engineering roles in industry. The research objective for the next five years is to adapt data management algorithms to the latest hardware and network technologies and create techniques that simplify the deployment of sensor systems for environmental and agricultural users. Database algorithms require optimization for devices with limited memory, and there is a need for more efficient algorithms for non-volatile memory devices such as NOR flash, which have unique storage properties. The latest, low-energy network protocols, such as LoRa, allow for innovative techniques for data management, as local data processing on the sensor can replace transmitting data over the network. We will build on our previously developed software and create new research to reduce the amount of data wirelessly transmitted by utilizing data compression, adaptive sampling, and machine learning techniques. These techniques will improve sensor efficiency by adaptively determining what data to store, transmit, and discard for best performance. This program will make fundamental contributions to understanding data processing with limited resources, release open source software to improve ease of use, and train a diverse group of five graduate and ten undergraduate HQP with skills in embedded software engineering. A primary use case is supporting time series data collected in agricultural applications. With increasing impacts of climate change and water scarcity, intelligent use of technology to improve sustainability, reduce water and fertilizer usage, and increase harvests is critical. A specific focus is continuing and expanding collaborations with the City of Kelowna, vineyard and agricultural producers in the Okanagan, and environmental agencies to deploy our research results to tackle key challenges of water management, climate monitoring, and agricultural productivity.