Recipient Organization
The Regents of University of California
200 University Office Building
Riverside,CA 92521
Performing Department
(N/A)
Non Technical Summary
Recent advances in robotics and automation, sensing, artificial intelligence (AI) and data sciences have created the possibility of gathering and assessing different crop-specific multi-modal field data at varying spatio-temporal scales and resolution. This has substantially increased the available amount of diverse multi-modal agriculture-relevant data. In turn, availability of more data has propelled forward AI-powered analysis, to provide more accurate and detailed information to various stakeholders (growers, farm advisors, agronomists), to improve several agricultural production outcomes like soil health mapping. Despite the availability of such large bodies of multi-modal data, the latency between data collections and follow-on data analysis remains high. Further, existing solutions to gather data are often limited as to the type of data that can be acquired (determined a-priori by the sensors a data collecting system [e.g., aerial robots considered herein] carries), and system configurability to facilitate collection of other types of data. Existing aerial robots are limited in terms of their reconfigurability and customizability, and seldom provide actionable data in real time. To address these limitations, this project proposes to leverage generative AI to develop and deploy customizable aerial robots that afford multi-scale and multi-type field data collections, and endow them with edge AI capabilities to ensure real-time data processing and analytics that offer on-demand crop information to stakeholders. The agronomic tasks considered include optimal harvesting time identification for grapes, efficient nutrient management for citrus, and soil health monitoring in both vineyards and citrus groves, over distinct climatic zones ([Southern] California, Florida, and [Northern] Greece). ?
Animal Health Component
10%
Research Effort Categories
Basic
70%
Applied
10%
Developmental
20%
Goals / Objectives
This project focuses on minimizing the latency in agricultural decision-making by harnessing generative AI, mobile robotics, and edge AI. The long-term goal of the proposed projectis to enable real-time adaptation of the data collection processes over multiple types of crops and minimize the time at which collected data can be processed to make decisions in support of critical agronomic tasks. To this end, the objectives of this project include:1) generative AI and robotics for scalable multi-modal data collections, and 2) edge AI for on-demand data analysis and real-time decision making, as well as a concurrent, multi-stage evaluation of the foundational and applied research methods studied in this project.We consider medium-scale (i.e. a few kg of weight) unmanned aerial vehicles (UAVs) that will be tasked to operate both over and inside/under the canopy in citrus groves and vineyards.The specific agronomic tasksconsidered include optimal harvesting time identification for grapes, efficient nutrient management for citrus, and soil health monitoring in both vineyards and citrus groves, over distinct climatic zones ([Southern] California, Florida, and [Northern] Greece). These considerations are motivated by our team's relevant background, previous efforts, and existing connections with the appropriate stakeholders and access to the necessary resources. Our project also seeks to integrateextension and education efforts throughout its duration.
Project Methods
Objective 1[Develop and deploy customizable UAVs that afford multi-scale and multi-type field data collections]:Weconsider three main activities under this objective: 1) establishing a mechanism for LLM-guided UAV specifications, 2) developing an automatic process to create synthetic agents and digital twins for rapid feasibility assessment of different UAV configurations, and 3) deploying and testing the AI-powered generated configurations into experimental fields.Taken together, these activities will yield a solution to develop and deploy customizable UAVs that can meet diverse stakeholder (grower, farm advisor, agronomist) needs while reducing the barrier for entry for UAV-based field data collections.Objective 2[Apply edge AI to reduce the latency between sensing and smart decision-making in agriculture]:Weconsider three main activities under this objective, structured in the form of a nature-language-based user need, as motivated in Objective 1 above.The three activities span three distinct agronomic tasks: 1) optimal harvesting time identification for grapes, 2) efficient nutrient management for citrus, and 3) soil health monitoring in both vineyards and citrusgroves. These will take place concurrently at different climatic areas:vineyards in California and Greece sites and citrus groves in California and Florida sites.This project will spend considerable effort on data collections to test and validate the research methods in the two objectives. These data collections will run alongside the two main objectives, and will consider three main activities: 1) standalone component testing and evaluation, 2) complete integrated system field testing where generative and edge AI will be brought together, and 3) assessment and comparison of integrated results in terms of current standard-of-practice processes in the three exemplary agronomic tasks considered herein (e.g., comparing against kriging interpolation for soil mapping over a vineyard or a citrus grove). This project will also contain an effort to perform data collections together with various stakeholders at small-scale yet fully integrated field experiments so at to test the technology readiness level of our methods, as the project matures.