Source: WASHINGTON STATE UNIVERSITY submitted to
AI INSTITUTE: AGRICULTURAL AI FOR TRANSFORMING WORKFORCE AND DECISION SUPPORT (AGAID)
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
EXTENDED
Funding Source
Reporting Frequency
Annual
Accession No.
1027037
Grant No.
2021-67021-35344
Project No.
WNP00874
Proposal No.
2021-07245
Multistate No.
(N/A)
Program Code
A7303
Project Start Date
Sep 1, 2021
Project End Date
Aug 31, 2024
Grant Year
2023
Project Director
Kalyanaraman, A.
Recipient Organization
WASHINGTON STATE UNIVERSITY
240 FRENCH ADMINISTRATION BLDG
PULLMAN,WA 99164-0001
Performing Department
School of EECS
Non Technical Summary
A growing world population will increase food demand while at the same time, agriculture faces complex challenges related to labor, water scarcity, weather events and climate change. The AgAID Institute will develop artificial intelligence (AI) solutions to help address these pressing challenges and spur the next agricultural revolution with the use of AI. More specifically, the Institute will build tools and workflows to help mitigate the effects of labor costs and shortages, and better manage regional resources such as water, despite climate uncertainties. The Institute will bring more data and science-guided information to the fingertips of agricultural workers to help them make better decisions.The Institute will emphasize solutions that can adapt to changing environments and climate, and amplify productivity through more efficient human and machine partnerships. Most importantly, however, the Institute's vision will be realized by making AI adoption its distinctive first principle. This means that that the people who will use the tools--the farmers, workers, and managers--will be deeply involved throughout all stages of the development process. The goal is to ensure the AI solutions are practical and add value, making them more likely to be used in dynamic real-world situations.Education and outreach are central to AgAID's activities not just to encourage AI adoption but also as a matter of justice. Raising AI skill levels and opening new career paths will increase compensation and improve quality of life for the agricultural workforce while attracting more people to agriculture and computing professions. This will be accomplished through K-12, college, and workforce training efforts carried out by the Institute's multi-disciplinary core members that include eight academic institutions and two tech companies as well as a range of public and private sector stakeholder groups. Partnerships with minority-serving institutions will engage Hispanic and Native American students. The Institute will work to develop inclusive AI interactions, such as bilingual and intuitive applications, to respond to the needs of a diverse workforce. By increasing AI education and closing skill gaps, the Institute aims to help transform this critical labor force and create new opportunities for computing and STEM majors.AgAID Institute's test cases will involve specialty crops such as apples, grapes, mint, and almonds that are grown nationwide and particularly in the Western U.S. These crops pose several agricultural grand challenges: they require intensive labor, need irrigation, and are heavily impacted by weather events and climate change. Specialty crops also account for 87% of the U.S. agricultural workforce. About 40% of these crops are perennial, requiring long-term management and planning. AI-based solutions that overcome these challenges for specialty crops will be more readily transferrable to other regions across the country and globally.If successful, the AgAID Institute's work will lead directly to the launch of several AI-powered systems used in agriculture in the near future. More importantly, the experience of successfully moving AI from the lab to operations will form a blueprint for rapidly addressing new agricultural challenges with AI. In this way, the AgAID Institute will provide long-term leadership in generating practical, real world solutions to address the complex web of challenges presented by a growing population and a changing climate.
Animal Health Component
0%
Research Effort Categories
Basic
30%
Applied
30%
Developmental
40%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
1110210205010%
1110210207010%
2041119208010%
4021119208010%
4021219208010%
6010210208010%
8031119208010%
8031219208010%
9030210208010%
9030430208010%
Goals / Objectives
The overarching goals of the AgAID institute are (a) to bring about a fundamental transformation in Ag decision support, farm operations and workforce development, using foundational developments in AI, and (b) to build a new coalition to create inclusive AI-Ag that is prepared and ready to take on future challenges of societal importance.Our institute's activities will be coordinated along nine interwoven dimensions or "thrusts" which include:Ag Thrusts:Water allocations intelligenceFarm operations intelligenceLabor intelligenceAI Thrusts:Modeling systems of knowns and unknownsMulti-scale decision supportInteractive human-AI workflowsDEMO FarmBroader Impact Thrusts:Adoption and technology transferEducation, extension and workforce developmentBroadening participationResearch thrusts and objectives:Ag Thrust 1. Water allocation intelligenceHypothesis: Addressing water-scarcity challenges will require new AI-enabled models of the coupled human and natural system, and AI-enabled decision support.Key objectives:WAI.obj1) To push the frontiers of hydrologic sciences by incorporating human-water nexus;WAI.obj2) To facilitate a shift from water "supply" to "availability" forecasts;WAI.obj3) To develop tools for optimal water allocation decisions..Use-cases:Irrigation schedulingPlantingInfrastructure planningAg Thrust 2. Farm operations intelligenceHypothesis: AI-enabled real-time, site-specific decision making can optimize farm resources while mitigating crop losses and improving produce quality.Key objectives:FOI.obj1) To develop site-specific models connecting accumulated management decisions to seasonal crop yield and quality outcomes;FOI.obj2) To construct a sensor-driven, adaptive real-time farm operations decision support framework.Use-cases:Frost mitigationDeficit irrigationHarvest managementAg Thrust 3. Labor intelligenceHypothesis: Challenges posed by increasing labor costs and a shortage in skilled labor workforce can be effectively addressed through human-machine partnerships.Key objectives:LI.obj1) To improve efficiency of existing field machines with AI;LI.obj2) To augment a less experienced workforce with intelligent machines;LI.obj3) To amplify more-experienced workers' productivity by training machines.Use-cases:Mechanical harvestingFlower thinningIntelligent pruningAI Thrust 1: Modeling Systems of Knowns and UnknownsKey objectives: AI.1.obj1) To design approaches to integrate simulators, data, and unknowns;AI.1.obj2) To design approaches for guided data collection and monitoring.AI Thrust 2: Multi-scale Decision SupportAI.2.obj1) To design approaches for site-specific real-time decision support;AI.2.obj2) To design approaches for long-horizon decision support.AI Thrust 3: Design of Interactive and Inclusive Human-AI WorkflowsAI.3.obj1) To design human-AI workflows for iterative human-centered design;AI.3.obj2) To design approaches for After-Action Review for AI;AI.3.obj3) To design approaches for unobtrusive instrumentation;AI.3.obj4) To design approaches for in-situ utility elicitation;AI.3.obj5) To design explainable interfaces;AI3.obj6) To design human-AI workflows with inclusive design.DEMO Farm (crosscutting research and broader impact)Key objectives:DEMO.obj1) To establish an interactive research platform for transdisciplinary teams to work side-by-side to better understand how AI technologies should be developed for different agricultural use cases;DEMO.obj2) To create an educational hub for students to gain hands-on experience;DEMO.obj3) To create an experiential learning site for key stakeholders including growers, field workers, educators, and technology providers.Broader impact thrusts and objectives:Adoption and technology transferKey objectives:TAT.obj1) To develop an ecosystem of intermediaries at the early stages of AgAID research and development to accelerate innovation feedback cycles and to facilitate technology adoption and transfer.TAT.obj1) To facilitate the creation of early adopter networks;TAT.obj2) To help create commercialization partners as part of an adoption flywheel;TAT.obj3) To identify areas where third party adoption could benefit user communities;TAT.obj4) To characterize how financial intermediation could support farmers at different scales.Education, extension and workforce developmentKey objectives:Ext.obj1) To facilitate a needs-driven co-development of inclusive and responsible AI tools via learning circles, resulting in tools with higher adoption likelihood, more adaptive to changing requirements, and better trust in AI for Ag participants;Ext.obj2) To jumpstart the adoption process by leveraging stakeholder relationships to recruit an early adopters' network;Ext.obj3) To support adoption amplification into user communities through training programs, and engagement with intermediaries to scale tech transfer and adoption.Ed.obj1) To increase STEM awareness at K-12 level, with a focus on middle and high school levels when students identify career preferences and often opt-out of STEM;Ed.obj2) To provide STEM retention pathways to students in junior colleges and 2-year associate degree programs which offer vital stepping stones to 4-year colleges, especially for lower income, underrepresented, and first generation students;Ed.obj3) To provide hands-on research experiences to undergraduate students for enriching their learning, influencing career choices and building occupational identity;Ed.obj4) To train and prepare graduate students and postdoctoral scholars who will form the next-generation workforce in research and technology development.Broadening participationKey objectives:BP.obj1) To explicitly engage Hispanic, Native American, and Women participants in all phases of AgAID implementation through an organizational cycle of identification, recruitment, training, practice, refinement, retention, and promotion;BP.obj2) To empower AgAID participants to engage in ideation, discovery, and implementation, while actively identifying and removing social barriers and biases;BP.obj3) To continuously assess and refine AgAID Diversity-Equity-Inclusion (DEI) activities at all educational levels (K-14, UG, G, PD), and to advance DEI within Ag-Tech workforce.
Project Methods
Approach:Tackling the grand challenges of 21st century agriculture will require fundamental shifts in the way we envision the role of AI technologies, and in the way we build AI systems. This shift is especially true for complex agricultural ecosystems such as in the Western U.S. The traditional approach toward development and deployment has been to view AI and technology designers as solution providers and the domain users as consumers. This monolithic view of producers and consumers, however, becomes grossly inadequate when brought to the fore in agriculture, which is a complex commercial multi-crop enterprise involving multiple stakeholders including the farmers (growers), farm laborers, consultants and technology service providers, state and regional policy makers, researchers and extension scientists, and students who form the future workforce. Therefore, for any AI-driven endeavor to succeed in this complex "Ag-sphere" there must be a strong alliance built between the AI designers and this broad range of stakeholders. Secondly, this AI designer - Ag stakeholder (people) alliance needs to be complemented by a strong AI technology - human factors (system) alliance - i.e., AI capabilities will need to include an inherent ability to integrate human input and account for human behavior. Humans can provide expert (scientific or in-field) guidance or be influential actors in complex dynamic processes (e.g. water use). Clearly, forging these two dimensions of alliances, people and system, will be beyond the scope of any specific research project or disciplinary silo, and warrants a transdisciplinary multi-party institute-scale effort - one that can propel AI developments throughout the Ag industry.To realize this vision, the AgAID institute will be built on the foundations of our partnerships between the team and our stakeholder groups, with AI, Ag, and humans as its three major intellectual pillars, and guided by three unifying principles that can be succinctly summarized as "Adopt-Adapt-Amplify". More specifically, we consider adoption as a first principle in AI design, adaptability to changing environments and scales, and amplification of human skills and machine efficiency - to be three important cross-cutting principles of design that will guide our approach to the core activities of our institute.Adoption as a first principle in AI design is removing barriers to AI technology adoption in Ag applications. This will be accomplished by: a) treating practical constraints and user considerations as central to the AI design process; and b) creating an environment of technology and knowledge co-production via proactive and continuous bidirectional engagement with the stakeholders.Adaptability to changing environments and scales is an ability that our approaches will inherently encode - to address the impacts of climate variability and weather fluctuations on agricultural productivity, and to provide decision support at multiple spatiotemporal scales of the Ag-sphere.Amplifying human skills and machine efficiency by augmenting automation with human skills and creating a close human-AI partnership will be critical to closing the gaps in workforce, while ensuring behavioral consistency and reduced uncertainty in decision support. Amplification will both enhance human skills and knowledge and improve machine efficiency, leading to a whole that is greater than the sum of its parts.Efforts:The AgAID Institute's activities will be coordinated along nine interwoven dimensions or "thrusts" that fall under three broad categories:[Ag thrusts]Water allocation intelligenceFarm operations intelligenceLabor intelligence[AI thrusts]Modeling systems of knowns and unknownsMulti-scale decision supportInteractive human-AI workflows[Broader impact thrusts]Adoption and technology transferEducation, extension and workforce developmentBroadening participationThe thrusts are structured in a way to create maximum synergy between the agricultural objectives through AI foundational advances. There are also strong ties between research and broader impact activities, as workforce training and stakeholder engagement are embedded within our research activities.Efforts will include (but not limited to): research training, research collaboration, extension and outreach, demonstration sessions and field trips, formal classroom instruction, laboratory instruction, practicum/internship experiences, certifications, and workshops and other experiential learning opportunities, and partnerships and nexus activities.Evaluation:Kansas State University's Office of Educational Innovation and Evaluation (OEIE) will serve as the AgAID evaluator. OEIE has extensive expertise in evaluation design, education research, quantitative and qualitative evaluation methods, robust data collection/analysis, and reporting, and has provided external evaluation for eighty-five NSF-funded projects. OEIE adheres to the American Evaluation Association's Guiding Principles for Evaluators and the Program Evaluation Standards of the Joint Committee on Standards for Educational Evaluation. The evaluation plan, includes formative evaluation to support project management and enable mid-course alterations, and summative evaluation to assess and document the overall impacts and outcomes. Evaluation activities are designed to gather data from the broad range of institute participants, and draw on both qualitative and quantitative methodologies and triangulate data when possible for more robust findings. OEIE will collaborate with the management team to finalize project milestones, progress metrics, and evaluation activities (e.g. surveys, interviews, and focus groups). Summaries will be submitted to project leadership to provide on-going feedback on project activities, integration of project components, and suggestions to maximize project success.

Progress 09/01/21 to 08/31/22

Outputs
Target Audience: Students (K-12, associate degree programs, undergraduate, graduate) Postdoctoral researchers Faculty at all AgAID member organizations External research collaborators in academia and industry Specialty crop growers Specialty crop agricultural farm managers and workers Crop commissions and crop consultants Water irrigation districts State departments of agriculture and ecology K-12 educators Associate degree program and undergraduate program educators IT industry partners AgTech and FarmTech industry partners Federal funding agencies including USDA NIFA and NSF Population groups: Underrepresented minority groups including (but not limited to) Hispanic/Latinx and Native American First generation college students Underrepresented genders Changes/Problems:?There has been no change in the scope or objectives of the project. However there have been a few changes in project personnel and collaborative arrangements, and a few challenges in student recruitment. As noted above none of these changes have impacted our overall scope or objectives. We report on these changes for the sake of completeness in reporting. Changes in collaborative arrangements: IBM Research was originally listed as a subawardee in the original proposal and at project initiation. However after discussions with our IBM Research collaborators (and some personnel changes at their end), it has now been mutually agreed that: IBM Research will still continue to collaborate with our team but not as a subawardee; instead they will serve as unfunded collaborators; the work that they were originally planning to cover on their end with respect to forecasting (relevant for the Water intelligence thrust) will now be performed by our team at WSU; IBM Research will still continue to provide access and support to their IBM PAIRS platform to ensure all objectives of our collaboration are met. This does not change the scope of the project. This change was relayed to the NIFA National Program Leader and has been approved. Changes in senior personnel: There have been a few changes in project personnel, in particular owing to faculty retirement or departures. Senior Personnel David Brown at WSU left academia for industry. His role in the original proposal was as a co-lead for the Farm intelligence thrust. He has been replaced by Lav Khot who is also Co-PI on the project. Co-PI Khot has also taken over as the AgWeatherNet Director, a role that was previously held by David Brown. Senior Personnel Claudio Stockle has retired and is no longer on the budget. He continues to serve as an unfunded collaborator in Emeritus Faculty status. His original role was in the Farm intelligence thrust and his responsibilities have been taken over by other senior personnel on the thrust (Markus Keller and Troy Peters). Senior Personnel Amit Dhingra has moved from WSU to Texas A&M University. Dhingra had only a minor role in technology adoption at a consulting capacity. He continues to hold an adjunct faculty role at WSU and will operate as an unpaid collaborator in the following years. COVID-19 related hiring and event challenges: There were a combination of hiring challenges and visa delays induced by COVID-19 on students and postdoctoral researchers. As the timing of the start of the award is not synchronized with the start of the semester, we have unexpended funds from year 1. However almost all of those activities have been initiated. Additionally, due to continued impacts on travel from the COVID-19 pandemic, some events were postponed or switched to virtual/hybrid events and as such, changed the original plans for events/travel in Year 1. What opportunities for training and professional development has the project provided?The AgAID Institute contributed to the training and professional development of a variety of individuals. A summary of the numbers by the different categories is presented below. Please note that the reported numbers represent a lower bound on the number of trainees. For names and areas of the individuals involved please refer to the detailed annual review report submitted to the cognizant NIFA National Program Leader. 4 postdoctoral scholars 43 graduate (PhD or MS) students 8 resident undergraduate students 14 summer undergraduate research interns How have the results been disseminated to communities of interest? Publications/papers: Abatzoglou, John, Daniel J. McEvoy, Nicholas J. Nauslar, Katherine C. Hegewisch, Justin L. Huntington. 2022. "Downscaled subseasonal fire danger forecast skill across the contiguous United States". Atmospheric Science Letters. (Submitted) Bertucci, Donald, Md Montaser Hamid, Yashwanthi Anand, Anita Ruangrotsakun, Delyar Tabatabai, Melissa Perez, and Minsuk Kahng. 2022. "DendroMap: Visual Exploration of Large-Scale Image Datasets for Machine Learning with Treemaps." IEEE Transactions on Visualization and Computer Graphics. arXiv. https://doi.org/10.48550/arXiv.2205.06935. (Accepted) Chatterjee, Amreeta, Lara Letaw, Rosalinda Garcia, Doshna Umma Reddy, Rudrajit Choudhuri, Sabyatha Sathish Kumar, Patricia Morreale, Anita Sarma, and Margaret Burnett. 2022. "Inclusivity Bugs in Online Courseware: A Field Study." In Proceedings of the 2022 ACM Conference on International Computing Education Research - Volume 1, 356-72. ICER '22. New York, NY, USA: Association for Computing Machinery. https://doi.org/10.1145/3501385.3543973. (Published) Dodge, Jonathan, Andrew A. Anderson, Matthew Olson, Rupika Dikkala, and Margaret Burnett. 2022. "How Do People Rank Multiple Mutant Agents?" In 27th International Conference on Intelligent User Interfaces, 191-211. IUI '22. New York, NY, USA: Association for Computing Machinery. https://doi.org/10.1145/3490099.3511115. (Published) Dodge, Jonathan, Roli Khanna, Jed Irvine, Kin-ho Lam, Theresa Mai, Zhengxian Lin, Nicholas Kiddle, et al. 2021. "After-Action Review for AI (AAR/AI)." ACM Transactions on Interactive Intelligent Systems 11 (3-4): 29:1-29:35. https://doi.org/10.1145/3453173. (Published) Guizani, Mariam, Igor Steinmacher, Jillian Emard, Abrar Fallatah, Margaret Burnett, and Anita Sarma. 2022. "How to Debug Inclusivity Bugs? A Debugging Process with Information Architecture." In 2022 IEEE/ACM 44th International Conference on Software Engineering: Software Engineering in Society (ICSE-SEIS), 90-101. IEEE Computer Society. https://doi.org/10.1109/ICSE-SEIS55304.2022.9794009. (Published) Homayouni, Taymaz, Akram Gholami, Arash Toudeshki, Leili Afsah-Hejri, and Reza Ehsani. 2022. "Estimation of Proper Shaking Parameters for Pistachio Trees Based on Their Trunk Size." Biosystems Engineering 216 (April): 121-31. https://doi.org/10.1016/j.biosystemseng.2022.02.008. (Published) Kalyanaraman, Ananth, Margaret Burnett, Alan Fern, Lav Khot, and Joshua Viers. 2022. "Special Report: The AgAID AI Institute for Transforming Workforce and Decision Support in Agriculture." Computers and Electronics in Agriculture 197 (June): 106944. https://doi.org/10.1016/j.compag.2022.106944. (Published) Khanna, Roli, Jonathan Dodge, Andrew Anderson, Rupika Dikkala, Jed Irvine, Zeyad Shureih, Kin-Ho Lam, et al. 2022. "Finding AI's Faults with AAR/AI: An Empirical Study." ACM Transactions on Interactive Intelligent Systems 12 (1): 1:1-1:33. https://doi.org/10.1145/3487065. (Published) Kokel, Harsha, Nikhilesh Prabhakar, Balaraman Ravindran, Erik Blasch, Prasad Tadepalli, and Sriraam Natarajan. 2022. "Hybrid Deep RePReL: Integrating Relational Planning and Reinforcement Learning for Information Fusion." In 2022 25th International Conference on Information Fusion (FUSION), 1-8. https://doi.org/10.23919/FUSION49751.2022.9841246. (Published) Koul, Anurag, Mariano Phielipp, and Alan Fern. 2022. "Offline Policy Comparison with Confidence: Benchmarks and Baselines." Transactions on Machine Learning Research. arXiv. https://doi.org/10.48550/arXiv.2205.10739. (Submitted) Saxena, Aseem, Paola Pesantez-Cabrera, Rohan Ballapragada, Kin-Ho Lam, Alan Fern, Markus Keller. 2023. " Grape Cold Hardiness Prediction via Multi-Task Learning". AAAI Conference on Artificial Intelligence. (Submitted) You, Alexander, Cindy Grimm, and Joseph R. Davidson. 2022. "Optical Flow-Based Branch Segmentation for Complex Orchard Environments." IEEE/RSJ International Conference on Intelligent Robots and Systems. arXiv. https://doi.org/10.48550/arXiv.2202.13050. (Accepted) You, Alexander, Nidhi Parayil, Josyula Gopala Krishna, Uddhav Bhattarai, Ranjan Sapkota, Dawood Ahmed, Matthew Whiting, Manoj Karkee, Cindy M. Grimm, and Joseph R. Davidson. 2022. "An Autonomous Robot for Pruning Modern, Planar Fruit Trees." IEEE Robotics and Automation Letters. arXiv. https://doi.org/10.48550/arXiv.2206.07201. (Submitted) Welankar, Sejal, Paola Pesantez-Cabrera, Ananth Kalyanaraman. 2022. "Extracting patterns in cold hardiness data using topological data analysis." Fourth International Workshop on Machine Learning for Cyber-Agricultural Systems (MLCAS2022). (Submitted) Saxena, Aseem, Paola Pesantez-Cabrera, Rohan Ballapragada, Kin-Ho Lam, Alan Fern, Markus Keller. 2022. "Grape Cold Hardiness Prediction via Multi-Task Learning." Fourth International Workshop on Machine Learning for Cyber-Agricultural Systems (MLCAS2022). (Submitted) Presentations: Ag2PI Field Day. Iowa State University INNOVATE'21. US Highbush Blueberry Council Integrative Precision Agriculture seminar. University of Georgia DnC2S meeting series. Oak Ridge National Laboratory North Central Washington Tree Fruit Granville Seminar. Pacific Northwest National Lab OSU Cherry Day 2022. OSU Extension Indigo Ag Science Seminar Series Tech Talk Tuesday School of EECS, Oregon State University International EUGAIN Summer Training School, Universita? della Svizzera italiana Summer Seminar Series IoT4Ag California Water Commission NCW TECH Alliance & North Central Educational Service District AI World Congress: AI x Agriculture: The New Era in Agriculture Other outreach events: NexTech Robotics 2021 SWE Conference Expanding your Horizons Tri-County Innovation Fair FLY CITRIS Drone Day Merced County Office of Education STEAM Festival Bobcat Summer STEM Academy What do you plan to do during the next reporting period to accomplish the goals?Farm Intelligence Projects F1: Cold hardiness prediction for frost mitigation Improve machine learning models and deploy them under practical settings Evaluate decision support interfaces and considerations F2: Weather data imputation and forecasting algorithms for site-specific modeling and use Design a spatio-temporal imputation mechanism that takes offline data from multiple weather stations in AgWeatherNet and imputes all missing sensor data. Design imputation approaches for predicting rest of the season weather data F3: Prediction of soil water content for deficit irrigation Develop and test machine learning models Compare and validate models with ground-truth obtained from Demo farm (smart vineyard) Obtain requirements for decision support interfaces F4: Policy comparison with uncertainty estimation using offline data Expand the benchmarks to include some Ag-inspired domains. Develop new offline policy comparison approaches that improve on the current metrics. Identify applications within the scope of AgAID decision support and apply F5: Building a topological data analytical framework for multi-cultivar spatio-temporal data Develop topological models and representation for frost mitigation and AWN data sets Apply and test topological models and incorporate with machine learning workflows Labor Intelligence Projects L1: Automated tree fruit pruning Identify pruning rules, workflow, and interface opportunities identified/developed to date Integrate developed rule set into robot arm pruning algorithms Conduct second phase human study to test/evaluate in the field L2: Automated flower thinning Understand human thinning decision making, developing thinning rules from formative studies Revise prototype, testing, crop yield, and quality analysis Improve deep learning models for individual flower detection L3: Mechanical harvesting in pistachios Analyze data collected from sensors for mechanical harvest Identify parameters for operator control and harvest tree manipulation Develop models for recommending optimal parameter settings for mechanical harvest Water Intelligence Projects W1: Streamflow forecasting Integrate domain knowledge into the graph neural network model for generating consistent predictions Develop methods for uncertainty quantification and large-scale empirical evaluation Expand into human influenced watersheds Explore alternative machine learning methodologies to model the forecast problem W2: Fallow prediction Improve the machine learning models for fallow prediction Expand to include information from California Conduct other extensions including estimating field age, prediction of fallowed fields under various drought scenarios, identifying changes in crop mixes, etc. W4: Detection of potential large-scale shifts in agricultural yields due to climate change Develop models to identify phase shift boundaries in the parameter space of VIC-CropSyst using active learning W5: IBM seasonal forecasts Conduct thorough comparative evaluation of accuracy of commercial seasonal forecast products available via the IBM PAIRS platform. Cross-cutting Research Projects Demo farm Continue and complete instrumentation of the two Demo farm sites (WSU Prosser and WVC) Use Demo farm sites for research thrust data collection and ground truthing Use Demo farm sites for training including boot camps and undergraduate research experience Human-AI workflows Develop and implement Ag use-case specific human-AI design workflows Apply and test inclusivity and interactability metrics Education Continue to train graduate and undergraduate students, and mentor postdoctoral scholars for research Conduct AgAID summer undergraduate research internships Conduct Digital Agathon Recruit students into PhD and MS degree programs in AI and in digital agriculture related topics Conduct MESA field trip and refine curriculum based on survey results Extension and workforce development Continue conducting learning circles for various research thrusts and to include new stakeholder interactions Continue extension activities for workforce training and broadening participation Broadening participation Pilot AgAID training/curriculum materials into HOEEP program to train current workforce Mentor students under LSAMP program Conduct bilingual training and outreach

Impacts
What was accomplished under these goals? Major activities: Launched the AgAID Institute in September 2021, including Institute website and project kickoff Initiated and launched multiple projects under research and multiple activities for broader impact and broadening participation Established communication channels and shared repositories and resources for collaboration Established project tracking and reporting systems Generated and conducted thrust level meetings and project level meetings for project initiation and implementation Worked with the external evaluation team to create and implement evaluation plan Received regular input and feedback from internal executive council for strategic decisions and Institute implementation Drafted and maintained Institute Strategic & Implementation Plan including authorship guidelines, EAB charter, IRB, and timelines and milestones Conducted the first cohort of summer undergraduate research interns Initiated stakeholder engagement and collaborations Initiated Institute level collaborations with external entities including industry and international partners Established the first cohort of External Advisory Board Specific objectives for ongoing work: In Year 1, we have initiated and have made progress across all research thrusts and broader impact thrusts. All projects are ongoing and in what follows, we list the set of objectives where work was initiated and where there has been significant progress. These objectives are listed by the different active projects ongoing under these different thrusts or across thrusts. Farm Intelligence Projects F1: Cold hardiness prediction for frost mitigation Develop machine learning models for grape cold hardiness prediction and compare those to existing models Develop and evaluate multi-task learning methods for cold hardiness of different grape cultivars F2: Weather data imputation and forecasting algorithms for site-specific modeling and use Develop and implement weather station correlation methodology for site-specific usage Exploration of robust weather data imputation methods and evaluate them on AgWeatherNet F3: Prediction of soil water content for deficit irrigation Determine variables of interest for prediction F4: Policy comparison with uncertainty estimation using offline data Development of methods for offline policy comparison F5: Building a topological data analytical framework for multi-cultivar spatio-temporal data Define topological data analysis workflows to capture multi-cultivar behavior in cold hardiness Labor Intelligence Projects L1: Automated tree fruit pruning Develop strategies and criteria for automated pruning through formative studies Investigate workflows and interfaces to effectively integrate robot pruning systems into existing farm workflows L2: Automated flower thinning Investigate/acquire manual operation and expert knowledge to develop strategies for flower thinning Investigate various deep learning models for detecting and locating flowers, and estimating flower cluster orientation and flower density L3: Mechanical harvesting in pistachios Installation of sensors to measure and map how individual trees are being shaken by different operators Water Intelligence Projects W1: Streamflow forecasting Combine the benefits of domain knowledge in the form of physics-based models and spatiotemporal graph neural networks to improve the overall prediction accuracy and uncertainty quantification W2: Fallow prediction Develop and evaluate ML models based on satellite time-series data for fallow prediction W3: Satellite data assimilation into the VIC model Improve the VIC hydrology model's ability to capture key snow and soil moisture processes by assimilating satellite imagery Cross-cutting Research Projects Demo farm Identify Demo farm sites at WSU (vineyard and apple orchard) and equip them with field sensors, imaging devices, drones, and field robots Identify a commercial orchard near WVC for a second Demo farm site, and install continuous plant monitoring sensors Human-AI workflows Design of interactive and inclusive human-AI workflows that allow for incorporation of human behavior and diversity of target groups Education Train graduate and undergraduate students, and mentor postdoctoral scholars for research Conduct AgAID summer undergraduate research internships and train 14 students Plan for Digital Agathon Initiate PhD and MS degree programs in AI Develop curricular materials for MESA field trip for middle school Extension and workforce development Implement extension activities to inform and contribute to research including learning circles and stakeholder interactions Initiate extension activities for workforce training and broadening participation Broadening participation Prepare and participate in Hispanic Orchard Employee Education Program Mentor students under LSAMP program Conduct bilingual training and outreach Work with MESA program to develop curriculum for computer science field trip to our Demo farm Significant results achieved: Ag-inspired AI research Development of cold hardiness predictive models for frost mitigation in grapes Formative studies to guide extraction of rules governing apple tree pruning Application of machine learning to develop predictive models for Ag land fallowing due to drought Foundational AI research Foundations for multi-task learning Developing the foundations for socio-economic mag Representation learning and science-guided ML for forecasting Cross-thrust integration Initiation of research projects and use-case prioritization Development of instrumented Demo farm at two sites for research and training Establish online portals and resources for intra-team training Establish data and modeling workflows and evaluation of computational platforms for research use Education Interdisciplinary training and tutorials in AI and Digital Ag Development of K-12 curricular materials for middle schoolers including NextTech robotics Research experience for undergraduates with inter-generational mentoring ?AI Expo aimed at educators at a STEM summit organized by NCW Tech Alliance and NCESD Initiation of PhD and MS in AI program at OSU Development of curricular materials for the first AgAID sponsored Digital AgAthon in partnership with Microsoft Extension and workforce development Learning circles and formative studies with Ag user groups, policy makers, and Hispanic community Preliminary studies for workforce training in orchard management Field days in orchard management (including in Spanish) Broadening participation and diversity Various K-12 training and outreach activities in Hispanic rural CA including coding camps and robotics programs Field day for K-12 MESA to Yakama Nation middle schoolers Recruitment and training of diverse groups as part of graduate and undergraduate research training Multi-organizational synergies/achievements Engagement with hitech and AgTech industry Outreach and presentations at various partner events including (but not limited to) Ag2PI, IoT4Ag, U.S. Highbush blueberry commission annual meeting, and SAS annual PD meeting Communication and securing infrastructure support from commissions Expansion of research into other specialty crops (berries, potatoes) Knowledge transfer Published or submitted at least 15 peer-reviewed papers and presented at numerous venues Training and orientation for AgAID Institute participants in intellectual property and commercialization Impact of the institute as a nexus point for collaborative efforts Establish a coordination and collaboration network with other sister NIFA AI Institutes (AIFARMS, AIFS, AIIRA) Establish communication with other NSF AI Institutes Initiate international collaborations (with Australia, Chile, and Netherlands) as part of the NSF-NIFA AI Institutes International Collaboration Supplements program Explore partnerships with the Cascadia Innovation Corridor and with Amazon AWS Key outcomes or other accomplishments realized: none to report yet

Publications

  • Type: Journal Articles Status: Under Review Year Published: 2022 Citation: Abatzoglou, John, Daniel J. McEvoy, Nicholas J. Nauslar, Katherine C. Hegewisch, Justin L. Huntington. 2022. Downscaled subseasonal fire danger forecast skill across the contiguous United States. Atmospheric Science Letters.
  • Type: Journal Articles Status: Accepted Year Published: 2022 Citation: Bertucci, Donald, Md Montaser Hamid, Yashwanthi Anand, Anita Ruangrotsakun, Delyar Tabatabai, Melissa Perez, and Minsuk Kahng. 2022. DendroMap: Visual Exploration of Large-Scale Image Datasets for Machine Learning with Treemaps. IEEE Transactions on Visualization and Computer Graphics. arXiv. https://doi.org/10.48550/arXiv.2205.06935.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Chatterjee, Amreeta, Lara Letaw, Rosalinda Garcia, Doshna Umma Reddy, Rudrajit Choudhuri, Sabyatha Sathish Kumar, Patricia Morreale, Anita Sarma, and Margaret Burnett. 2022. Inclusivity Bugs in Online Courseware: A Field Study. In Proceedings of the 2022 ACM Conference on International Computing Education Research - Volume 1, 35672. ICER 22. New York, NY, USA: Association for Computing Machinery. https://doi.org/10.1145/3501385.3543973.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Dodge, Jonathan, Andrew A. Anderson, Matthew Olson, Rupika Dikkala, and Margaret Burnett. 2022. How Do People Rank Multiple Mutant Agents? In 27th International Conference on Intelligent User Interfaces, 191211. IUI 22. New York, NY, USA: Association for Computing Machinery. https://doi.org/10.1145/3490099.3511115.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Dodge, Jonathan, Roli Khanna, Jed Irvine, Kin-ho Lam, Theresa Mai, Zhengxian Lin, Nicholas Kiddle, et al. 2021. After-Action Review for AI (AAR/AI). ACM Transactions on Interactive Intelligent Systems 11 (34): 29:1-29:35. https://doi.org/10.1145/3453173.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Guizani, Mariam, Igor Steinmacher, Jillian Emard, Abrar Fallatah, Margaret Burnett, and Anita Sarma. 2022. How to Debug Inclusivity Bugs? A Debugging Process with Information Architecture. In 2022 IEEE/ACM 44th International Conference on Software Engineering: Software Engineering in Society (ICSE-SEIS), 90101. IEEE Computer Society. https://doi.org/10.1109/ICSE-SEIS55304.2022.9794009
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Homayouni, Taymaz, Akram Gholami, Arash Toudeshki, Leili Afsah-Hejri, and Reza Ehsani. 2022. Estimation of Proper Shaking Parameters for Pistachio Trees Based on Their Trunk Size. Biosystems Engineering 216 (April): 12131. https://doi.org/10.1016/j.biosystemseng.2022.02.008.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Kalyanaraman, Ananth, Margaret Burnett, Alan Fern, Lav Khot, and Joshua Viers. 2022. Special Report: The AgAID AI Institute for Transforming Workforce and Decision Support in Agriculture. Computers and Electronics in Agriculture 197 (June): 106944. https://doi.org/10.1016/j.compag.2022.106944.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Khanna, Roli, Jonathan Dodge, Andrew Anderson, Rupika Dikkala, Jed Irvine, Zeyad Shureih, Kin-Ho Lam, et al. 2022. Finding AIs Faults with AAR/AI: An Empirical Study. ACM Transactions on Interactive Intelligent Systems 12 (1): 1:1-1:33. https://doi.org/10.1145/3487065.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Kokel, Harsha, Nikhilesh Prabhakar, Balaraman Ravindran, Erik Blasch, Prasad Tadepalli, and Sriraam Natarajan. 2022. Hybrid Deep RePReL: Integrating Relational Planning and Reinforcement Learning for Information Fusion. In 2022 25th International Conference on Information Fusion (FUSION), 18. https://doi.org/10.23919/FUSION49751.2022.9841246.
  • Type: Journal Articles Status: Under Review Year Published: 2022 Citation: Koul, Anurag, Mariano Phielipp, and Alan Fern. 2022. Offline Policy Comparison with Confidence: Benchmarks and Baselines. Transactions on Machine Learning Research. arXiv. https://doi.org/10.48550/arXiv.2205.10739.
  • Type: Conference Papers and Presentations Status: Under Review Year Published: 2023 Citation: Saxena, Aseem, Paola Pesantez-Cabrera, Rohan Ballapragada, Kin-Ho Lam, Alan Fern, Markus Keller. 2023. " Grape Cold Hardiness Prediction via Multi-Task Learning". AAAI Conference on Artificial Intelligence.
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2022 Citation: You, Alexander, Cindy Grimm, and Joseph R. Davidson. 2022. Optical Flow-Based Branch Segmentation for Complex Orchard Environments. IEEE/RSJ International Conference on Intelligent Robots and Systems. arXiv. https://doi.org/10.48550/arXiv.2202.13050.
  • Type: Journal Articles Status: Under Review Year Published: 2022 Citation: You, Alexander, Nidhi Parayil, Josyula Gopala Krishna, Uddhav Bhattarai, Ranjan Sapkota, Dawood Ahmed, Matthew Whiting, Manoj Karkee, Cindy M. Grimm, and Joseph R. Davidson. 2022. An Autonomous Robot for Pruning Modern, Planar Fruit Trees. IEEE Robotics and Automation Letters. arXiv. https://doi.org/10.48550/arXiv.2206.07201.
  • Type: Conference Papers and Presentations Status: Under Review Year Published: 2022 Citation: Welankar, Sejal, Paola Pesantez-Cabrera, Ananth Kalyanaraman. 2022. Extracting patterns in cold hardiness data using topological data analysis. Fourth International Workshop on Machine Learning for Cyber-Agricultural Systems (MLCAS2022).
  • Type: Conference Papers and Presentations Status: Under Review Year Published: 2022 Citation: Saxena, Aseem, Paola Pesantez-Cabrera, Rohan Ballapragada, Kin-Ho Lam, Alan Fern, Markus Keller. 2022. Grape Cold Hardiness Prediction via Multi-Task Learning. Fourth International Workshop on Machine Learning for Cyber-Agricultural Systems (MLCAS2022).