Recipient Organization
UNIVERSITY OF CALIFORNIA, DAVIS
410 MRAK HALL
DAVIS,CA 95616-8671
Performing Department
(N/A)
Non Technical Summary
AI-enabled technology for agriculture and food systems are advancing at a rapid pace from crop breeding and production through food processing and consumption. A key barrier in the development of AI-enabled innovations in food and agriculture is the availability of high-quality, large-scale, relevant datasets to train deep AI models. Collecting this data and conducting ground truth observation is costly, often privately held, and may take years and global travel to obtain. Accurate, tested, and robust datasets and machine learning (ML) models derived from, and supporting end-to-end food system components are needed. Knowledge of how to curate, retrieve, and use these models and data are also critical. The availability and knowledgeable implementation of these models and data can dramatically lower the cost and time to effect change in food system informatics.Developing such data and models for food and agriculture is challenging as they represent complex interactions between biological systems, machinery, sensors, and humans4. They require cross-disciplinary expertise from plant and food science combined with engineering and computer science, among other disciplines. Universities across the world are making progress in the development of ML models and data sets for AI development in food and agriculture, with varying focus areas and approaches.Building upon the significant progress of the current award, AIFS will engage a key international alliance, Data Driven AgriFood Future (DDAF) and its constituents in a multi-pronged collaborative effort involving establishing common ground for AI innovation, articulating the problems to be solved, engaging in specific research related to these problems, and participating virtually in international workshops to allow all alliance members to share and learn from each other, and to chart a path forward.AIFS Graduate students will travel to Waginengen University and Research (WUR) in Netherlands and Fraunhofer ISE in Germany for this collaborative research, and will be guided by both faculty/staff at those institutions and from their participating AIFS faculty.
Animal Health Component
25%
Research Effort Categories
Basic
50%
Applied
25%
Developmental
25%
Goals / Objectives
AIFS will lead a global initiative as part of the DDAF alliance entitledAIData forAgrifoodPlatformTechnology (ADAPT) that will drive AI for agrifood international collaboration by creating a common platform for open technology sharing, starting with three project thrust areas and two partners.3.1 Thrust 1: AI to accelerate crop development for controlled environments.A current barrier to both new crop development and precision crop production in controlled environment agriculture (CEA), including greenhouse production using sunlight and sunless vertical farming under lights, is the ability to quickly and accurately analyze plant phenotypes and their meaning7. Phenotyping is an important approach in analyzing and predicting crop phenotype (interaction of genetics and environment) for development and selection of new genetics for crops.Facilities such as the newly constructed phenotyping facility at Wageningen University and Research (WUR) provide reference environments for cultivation from new seeds and analysis of critical selection features used as predictors of the phenotype for new seed genetics. Researchers record large amounts of data during phenotyping but much has limited utility currently. An emerging trend is to decrease collection of data to simplify/streamline the analysis task. The ability to better utilize phenotyping data in CEA has important use-inspired potential. Seeds for new, high-performance CEA crop selections are of great value - prices of $2.50 per seed are common.AIFS will work with the WUR Netherlands Plant Eco-Phenotyping Centre (NPEC) to develop lower cost tools to support genotype-phenotype associations that use AI/ML models and lower cost hardware, baselined against the NPEC phenotyping platform to advance AI/ML modeling to lower the cost and time of high-throughput phenotyping.The goals of this thrust are to:Co-develop AI/ML and other data science tools with our international collaborator(s) to accelerate crop development using currently underutilized phenotyping data in CEA.Develop lower-cost approaches to highly accurate phenotyping using readily-available equipment with newly-developed data and software tools.3.Catalyze a broader international collaboration among leading research institutions as well as private sector companies and government agencies to inform use-inspired research and begin translation activities.3.2 Thrust 2: Digital Twin to Advance Agrophotovoltaics (APV)Agriculture accounts for almost 70% of the world's water use, which is expected toincrease even more in the coming years9. Irrigation of large farm fields requires extensive energy, particularly for pumping water. Water use and energy generated from fossil fuel need to be reduced for a sustainable food system10. APV systems, which co-develop the same area of land for both solar photovoltaic power and agriculture, provide great opportunities to achieve this objective.The primary goals of this thrust are to:Co-develop with our international collaborator easy-to-use data science and simulation tools that are inexpensively accessible to a wide range of persons involved in APV systems, equitably and fairly.2.Share this gained knowledge and modeling among collaborators via virtual presentation, preferably a hybrid or virtual conference, and to look for next steps in collaboration to move this field forward in scope and translation to industry.This will accelerate the development of new technologies to optimize agricultural efficiency, quality, safety, water distribution and energy management, enabled by data science, high-performance computing, and systems engineering.A central component of this activity is the "Digital Twin" paradigm of physical reality, i.e., digital replicas of complex systems that can then be inexpensively and safely updated, improved and optimized in a virtual setting and deployed in the physical world, reducing the potential costs of experiments and accelerating development of new technologies11.By creating digital twins of complex, symbiotic APV systems, the project will safely and efficiently simulate, improve, and optimize agriculture, careful water use, and integrated solar energy in virtual settings, before deploying them in the physical world. An important aspect of this activity is to develop a fully functional APV test-bed where data can be extracted, processed, controlled, and consequent actions implemented through a digital twin.3.3 Thrust 3: AI to Improve Food Safety in Controlled Environment AgricultureFoodborne illness affects 48 million people in the U.S. each year12. Chemical and microbial safety of food is a critical societal challenge, compounded by the complexity of the supply chain, with the potential to introduce contamination at various points.Contamination can occur during production, harvest, processing, storage, distribution, or at retail/preparation12. Controlled environment agriculture, both greenhouse (sun) and vertical farm (sunless) production are perceived to have fewer food safety concerns than field-grown produce due to tighter control and filtration of water supplies, and isolation from soil and wild animals.However, pathogens can still be introduced into CEA production systems through water, substrates, and human contact. A 2021 salmonella outbreak at greenhouse producer BrightFarms caused the recall of lettuce products in four U.S. states and a shutdown of operations for three months13.CEA environments provide an opportunity to develop and apply food safety technologies that are common in food processing environments, and cutting edge techniques using sensors and AI/ML models may detect and prevent pathogens in real-time.AIFS will work with the WUR Netherlands Greenhouse Technology and Food Safety Research teams to develop digital twin models for the greenhouse and vertical farming environments that focus on real-time pathogen detection and prevention. The approach will leverage work done for food processing environments, including data from fresh produce industries and public data sets in the U.S. as well as from the EU. Instrumented greenhouses and vertical farm environments at both Cornell University and WUR in the Netherlands will provide the basis for digital twin model development.The goals of this thrust include:Co-develop AI/ML and other data science tools with WUR to improve food safety inCEA.Develop lower-cost approaches to improving real-time pathogen detection and prevention using readily-available equipment with newly-developed data and software tools.3.Catalyze a broader international collaboration among leading research institutions as well as private sector companies and government agencies to inform use-inspired research and begin research translation activities.
Project Methods
Methods for thrust 1 (phenotyping):The approach of this thrust will be:Develop methods to extract current high-value genetic indicators in the phenotype.Develop AI and data science tools to identify new genetic indicators that compliment current knowledge.Expand data collection and analysis methods with new sensors and systems integration.Designate a graduate student researcher to facilitate work between AIFS and international collaborators5.Hold regular virtual meetings between AIFS and WUR researchers, facilitated by the graduate student employee. Travel to the EU for in-person experimentation at the NPEC facility in the Netherlands and then following year 1 to share findings, demonstrate tools, and advance further collaboration.Methods for thrust 2 (Agrophotovoltaics):Identify, in conjunction with the collaborative partner, key data elements and challenges with existing data sets involved in advancing the robustness of APV data models.Build data sets based on US and international crop and photovoltaic site sources.Develop digital twin models based on these data sets.Designate a graduate student researcher to facilitate work between AIFS and international collaborators5.After year 1 or during, depending on timing, share findings, demonstrate tools, and advance further collaboration via virtual presentation.Methods for thrust 3 (safety):Leverage existing data sets and create digital twins of research CEA facilities in the U.S. and EU to develop new models to detect and prevent pathogens in CEA environments.Instrument research CEA facilities for pathogen detection to expand data collection and analysis methods with new sensors and systems integration.Identify publicly available data sets for microbial ecology to be integrated with chemometric and physical measurements from the CEA harvesting and processingDesignate graduate student researcher to facilitate work between AIFS and international collaborators5.Hold regular virtual meetings between AIFS and WUR researchers, facilitated by the graduate student employee. Travel to the EU for in-person experimentation at the WUR Greenhouse Horticulture facility in Bleiswijk, Netherlands and then following year 1 to share findings, demonstrate tools, and advance further collaboration.