IoT and edge computing technologies for vertical farming from seed to harvesting

Abstract

With the ever growing global population, a key challenge of the future will be providing enough food for people. One proposed solution to this is vertical farming, which is the practice of growing crops vertically, above each other in order to increase total crop yield for a given space. Supported by IoT technologies, such as sensors and artificial lighting, it allows farmers to utilize spaces that previously did not provide suitable growing conditions for crops, such as underground car parks, or abandoned underground tunnels. By utilizing developments in IoT technologies, and mechanisms for efficient management of large volumes of IoT-generated data, such as stream processing, it is now possible to measure and analyse the vast amount of parameters that influence plant growth, such as humidity or temperature. This makes it possible to facilitate continuous monitoring and decision making at runtime towards a fully autonomic management of vertical farming processes. Importantly, these technologies can provide support for long-term storage and analysis of such data, in order to create new domain knowledge, such as by developing databases that contain optimal growth parameters for a majority of edible plants.In this thesis we lay the groundwork for such mechanisms. We analyse the software design requirements for a system that can gather such information, as well as analyse various network topologies and deployment strategies for it. In addition, we show an implementation that already can be used to gather important data.Our software architecture lays a focus on extensibility and allows others to easily build upon it. Last, we perform extensive testbed experiments to evaluate the performance of our scheme and to demonstrate its feasibility to operate in diverse deployment scenarios and network architectures and topologies. Aiming to identify the performance limits of low-cost edge computing equipment, we deploy the components of our architecture on low-end single board computers and show them to be able to support monitoring workloads that correspond to thousands of sensing devices. We further explore experimentally the performance and design implications of using modern Low-Power Wide Area Networking technologies, and in particular LoRaWAN, to collect monitoring data from remote vertical farms with minimal network infrastructure requirements. We show that while deploying and managing a full private LPWAN over edge computing devices may have significant effects on the overall system's workload processing capacity, it is still possible to support large-scale vertical farm installations.