Source: IOWA STATE UNIVERSITY submitted to
AI INSTITUTE: AIIRA: AI INSTITUTE FOR RESILIENT AGRICULTURE
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
EXTENDED
Funding Source
Reporting Frequency
Annual
Accession No.
1027030
Grant No.
2021-67021-35329
Project No.
IOWW-2021-07266
Proposal No.
2021-07266
Multistate No.
(N/A)
Program Code
A7303
Project Start Date
Sep 1, 2021
Project End Date
Aug 31, 2027
Grant Year
2024
Project Director
Ganapathysubramanian, B.
Recipient Organization
IOWA STATE UNIVERSITY
2229 Lincoln Way
AMES,IA 50011
Performing Department
Mechanical Engineering
Non Technical Summary
Our planet faces a daunting challenge: By the end of the century, world population will increase by 45%, cropland will decrease by 20% and our climate will become increasingly variable, threatening crops and putting communities at risk. We need to increase agricultural productivity by 70% to meet our growing food security needs - a challenge we are not able to meet under our current rate of progress.Now imagine a truly game-changing technology that can greatly accelerate this progress. It already exists in the form of artificial intelligence (AI). Using advanced sensor technology, scientists can create digital twins - virtual simulations that mimic real-world plants, crops and farms. For every year of biological data, digital twin-based AI systems can create hundreds of reality-based simulations that can:Streamline and revolutionize plant breeding to help scientists develop improved crop varieties that can better withstand environmental, pest and disease challenges while delivering higher yields and quality.Help farmers and their advisors adopt improved farming techniques and technologies that can boost their profits and help improve the long-term care of their critical land and soil resources.Provide governments with the insight they need to encourage and incentivize adoption of policies and practices that deliver the most benefit with the least environmental cost.Give agricultural companies the data and knowledge needed to develop more effective precision management systems and improved plant varieties that thrive with less water, fertilizer and pesticides.Drive economic development across the rural landscape through AI-inspired ventures.The leaders of the AI Institute for Resilient Agriculture (AIIRA) believe these breakthroughs - and more - can be a reality in the very near future. The Institute is bringing together AI experts with plant breeders, agronomists, geneticists and social scientists to accelerate the adaptation and use of AI-based technologies to transform agriculture to meet the needs of our world's growing population and increasingly climate-challenged food systems.
Animal Health Component
0%
Research Effort Categories
Basic
50%
Applied
40%
Developmental
10%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
2121820208010%
2121820108110%
2051820208010%
2031820208010%
2031510208010%
2051510208010%
2011510208010%
2011510108110%
2062499208010%
9032499208010%
Goals / Objectives
The project has the following goals and associated objectives:Goal 1: Build plant and field scale predictive models through foundational AI advances. The specific objectives related to this project are:Objective 1: Build AI algorithms and perform associated theoretical analyses for (a) optimal sensor placement, (b) remote and proximal perception, and (c) programmable sensing for ag applications.Objective 2: AI theory and algorithms that encode contextual/domain knowledge into AI constructs. These context-aware models fuse multi-scale, multimodal data that will result in accurate predictive models for biological entities.Objective 3: Develop and deploy novel AI planning and control capabilities so robots can intelligently execute in-field actions like data collection and dexterous manipulation of plants and their environment. Goal 2: Deploy plant and field scale predictive models for breeding and crop production applications. The specific objectives related to this goal are:Objective 4: AI theory and algorithms that enable in silico breeding, and AI-enabled decision-support tools that reduce resource utilization while managing risk. This includes AI-enabled optimization in high-dimensional spaces for ideotype design, partial data-based optimization with guarantees, reinforcement learning for coupled in silico and conventional breeding.Objective 5: Extend predictive model based decision-making to field-level optimization of production tasks and develop distributed AI approaches to enable multiple mobile agents to collect data, take action, and improve models under dynamic conditions at different spatiotemporal scales.Goal 3: Understanding and resolving social barriers to, and AI innovations for adoption of the AI technology in the agricultural ecosystemObjective 6: Identify social, behavioral, and business catalysts and barriers to acceptance and adoption of AI technology by various stakeholders in the ag ecosystem.Objective 7: Develop foundational algorithms and theory to increase trustworthiness of AI tools, specifically to enhance the acceptance and adoption of AI-enabled technologies by the ag community. Goal 4: Create a diverse, AI-aware agricultural workforce and serve as a nexus for AI-in-Ag developmentsObjective 8: Develop talent and skills for a highly competent next generation AI workforce, including activities for graduate students; undergraduate students; and continual learners with diverse backgrounds,learning levels, and fields of expertiseObjective 9: Broadening participation of women, Hispanics, and Native Americans in AI using evidence based strategies across multiple age groups.Objective 9: Team science based collaboration and sustained knowledge transfer activities.
Project Methods
Effort: Our efforts will include:Principled approaches for sensor allocation and placement, along with advances in context-aware 3D sensing, will provide layered data to feed the digital twinBuild, train, and validate a new family of deep generative models that are explicitly physics aware and that encode biophysical and phenomenological constraints, physiological crop models, bio-eco-hydrological process models, and network associations within the training processExplore the use of forceful interaction to inform the digital twin, model-based strategies for manipulation of flexible and granular material with soft manipulators, and AI-driven design of robotic mechanisms well suited to a given dexterous task. A systems integration component will support in-field testing and application of developed AI advancesAI-enabled optimization in high-dimensional spaces for ideotype design, partial data-based optimization with guarantees, reinforcement learning for coupled in silico and conventional breeding.New algorithms and theory for distributed optimization and learning to solve various multi-agent reasoning problems. Important practical constraints, such as limited communication and privacy awareness, will drive the AI innovationsIdentify social, behavioral, and business catalysts and barriers to acceptance and adoption of AI technology by various stakeholders in the ag ecosystem. Use a combination of quantitative and qualitative approaches: community of practice, focus groups, online surveys, participatory workshops and economic experiments, semi-structured interviews, and the Delphi method.Define, formulate, and develop algorithms and metrics of interpretability and robustness for AI frameworks; develop principled approaches to incorporate knowledge of domain expertsEvaluation: AI research will be evaluated via metrics including computational efficiency, sample efficiency, generalization to model fidelity, and effectiveness of data collection. Application-oriented impacts to genetics, breeding, and production will be evaluated through cross-cutting field trials that integrate and deploy AI advances and measure their efficacy for multiple crops and growing regions. Social impact will be measured by a community of practice, regional and national surveys, focus groups, and field experiments, and the evaluation will examine the dissemination of findings on catalysts and barriers to AI adoption to stakeholders. Formal and informal education and workforce development modules will be formatively and summatively accessed. Outreach will be assessed by the number of stakeholders engaged and changes in their AI-ag knowledge and attitudes

Progress 09/01/22 to 08/31/23

Outputs
Target Audience:AIIRA reaches a broad and diverse audience of USDA/NIFA stakeholders through our research and outreach activities. • K-12 students (middle and high school) including students of color and students from economically disadvantaged backgrounds • Graduate students and postdoctoral fellows • Undergraduate and community college students including women, native american and hispanic populations • Farmers and commodity groups including Indigenous Farmers • Extension and teaching professionals, Ag professionals • USDA ARS scientists • Industry Changes/Problems:Other personnel: Aaron Presholt was serving as an interim PI from Iowa Soybean Association. He was replaced by Matthew Carroll in Spring 2023. Dr. Carroll is very familiar with many aspects of the AIIRA project as former graduate student within AIIRA. Budget: AIIRA's year 2 funding was not received until December 2022 even thought the project year began in September. As we were underspent in year 1, we were able to obtain a no cost extension to cover student stipends and tuition until the year 2 funds arrived, but it was an unexpected challenge to maintain continuous support during the delay. What opportunities for training and professional development has the project provided?AIIRA's education and workforce development opportunities aim to develop talent and skills for a highly competent next generation AI workforce which includes activities for graduate and undergraduate students, and continual learners of all ages. A diverse team from across AIIRA's collaborative institutions aim to leverage existing partnerships to create new learning opportunities and expand existing programs aimed at advancing knowledge and education in AI in Ag. University and college: AIIRA personnel directly mentored 7 postdocs, 84 graduate students, and 12 undergraduates across 8 institutions. Mentoring consisted of both formal course work, training and research assistantship, as well as informal networking across institutions. We briefly detail selected activities below, with additional details in the full report to the cognizant program manager. Undergrad/grad education: 1) We made progress planning for undergraduate programming in AI-in-Ag (an AIIRA Signature Activity). AI for Robotics course was developed and taught at CMU. 2) GMU has prepared a case study course on ethics of use in AI for Ag. 3) ISU has developed an undergraduate minor in engineering called Cyber-Physical Systems (an AIIRA Signature Activity) with modules planned for AI-in-Ag with a capstone course that serves as a template AIIRA can use for a graduate minor. 4) We are planning for Ag-AI content in the recently offered Data Science minor at ISU, focused on undergrads from Ag and Life Sciences. 5) ISU has launched an entrepreneurial marketing course wherein students are working on marketing projects for Cyverse and Big Data in a Box. Founders of these entities have given guest talks to students. In addition, non-profit organizations in the ag and food security space such as Cultivating Hope Farms (a farm for persons with disabilities run by a woman farmer) and E-Feed Hungers (a social enterprise linking food banks with beneficiaries) are also collaborating with the entrepreneurial marketing students. 7) Chinmay Hedge served as a GSTEM (formerly Girls in Stem) mentor at NYU Tandon in the summer of 2023, 8) ISU's AGRON 279 Field Exploration in Agronomy demonstrated to undergraduate students on what a self-pollinated species breeding program looks like with a focus on technology for farmers like the insect app, and 9) ISU's AGRON 183 highlighted the applications of UAVs and Rovers in agricultural studies. Workforce development via Ag-AI related training opportunities: 1) We offered application development exploring various AI and software tools with several honors freshmen, one of which built a prototype for their summer research. 2) Provided opportunities for AI, digital transformation and digital agriculture research involving one honors marketing undergraduate Grace Pearson and an MBA graduate student Amrutha Kandhare (both women), 3) We mentored students on entrepreneurial marketing projects in digital agriculture, agricultural and applied economics as it relates to ethical and social barriers of AI-in-Ag as well as in Computer Science and Engineering introducing them to digital agriculture research. 4) Four graduate and two undergraduate students worked on different aspects of AI, AI Ethics, and AI ethics and agriculture related projects resulting in case studies for instruction and implemented in an undergraduate course in Fall 2022. 5) The development of ML algorithms by two undergraduate students, who took into account interaction with humans and the capabilities of human feedback. 6) Exposed a sustainable agriculture & sociology PhD student to the social dimensions of AI-in-Ag literature and research and provided opportunities to present their work at research group meetings or to the research community. 7) Several engineering undergraduates were exposed to AI-in-Ag research opportunities. 8) Generated engineering capstone design teams. 9) We provided professional presentation coaching to five AIIRA graduate students (3 women) in preparation for their presentations to our USDA cognizant program leader at our annual review. 10) Currently working with an under-graduate research assistant Yufei Zhang and a graduate assistant Prathiksha Ravi Krishnan (both women) on consumer research on AI, sustainability and ag. Workforce development via connecting with stakeholders: AIIRA completed activities to generate a new AI-in-Ag workforce. 1) Efforts were focused on identifying means to connect with farmers and growers to demonstrate AI driven tools and technologies as well as means to showcase AI driven technologies to the broader agricultural community. 2) AIIRA presence at the 2022 Farm Progress Show in Boone, IA, with over 10,000 visitors to the tent. 3) The Data Science for the Public Good, Wholesale Local Food Benchmarking Project sponsored by AIIRA Expanded on the previous year's project. The Data-drive Insights for Local Food Markets: AI, Pricing, and Crop Flow project brought together one undergraduate student, three graduate students (with degrees in Management Information Systems, Data Science, History, Industrial Engineering and AI), extension professionals, and several support faculty to enhance local food markets by providing valuable insights and optimizing the crop flow. Results benefit farmers by creating the ability to provide personalized recommendations regarding which crops to grow based on their respective locations. Workforce development via community building activities: We organized several workshops to disseminate research and education advances to various stakeholders including: 1) International Workshop on Machine Learning for Cyber Ag systems (MLCAS) held July 3-5 in Japan, 2) Translational AI Industry and Academia Symposium in June 2023 connected Industry representatives from across Iowa to AIIRA AI experts to discuss ways in which Iowa State University and Industry can expedite adoption of Artificial Intelligence, Machine Learning, and Big Data Analytics. AIIRA met with over 20 organizations interested in expanding workforce development across Iowa. Pre-college education: AIIRA's pre-college education efforts focused on leveraging existing partnerships at CMU with a goal to develop a new curriculum in Pittsburgh, PA through the Urban Robotony (an AIIRA Signature Activity) workshop for middle schoolers. As a signature AIIRA effort, the program is a hands-on program targeting middle school students that combines AI, robotics, and plant science. The second year of the program continued on the success of the first with the second iteration of Urban Robotony occurring in July 2023. 2) ISU's summer internships resulted in three high school students learning machine learning coding as it relates to field corn and common agricultural insects. AIIRA sponsored a unique informal learning opportunity to engage with undergraduate students outside of the AI and Agriculture frames. A. Johri (GMU) and an AIIRA graduate student implemented the AIIRA Writing Contest asking applications to either write a letter imaging what a meal could be in 2050 and the technology used to produce, transport, and prepare that meal, or create a menu or recipe that describes the sourcing and production of ingredients. Both scenarios encourage the writer to consider ethical dilemmas involved. The contest received 160 entries and awarded 3 overall prize winners, 3 separate winners to Youngstown University students and 4 honorable mentions. The contest will continue in Fall 2023 and be expanded to include high school entrants. How have the results been disseminated to communities of interest?AIIRA results are disseminated through collaboration and knowledge transfer workshops and training, sharing of benchmark datasets, and participation in field days to engage directly with stakeholders. Multidisciplinary Integration and Community Building. AIIRA has worked towards creating a learning community to support a collaborative research and training environment across the transdisciplinary boundaries of team members. We are administratively housed within the Center for Translational AI (TrAC) at ISU specifically to facilitate collaboration and knowledge transfer activities. With the goal of providing the means to support collaboration and sharing of knowledge, and using team science-based approaches, many workshops and training have been established or shared across the team and AI-in-Ag community. AIIRA sponsored workshops and training: AIIRA sponsored, developed, and implemented several workshops and training opportunities to AIIRA stakeholders. 1) TrAC Journal Club included AIIRA graduate students who participated in 10 sessions offered in Fall 2022 through Spring 2023. 2) TrAC Seminar Series sponsored by AIIRA presented to a wide audience across disciplines. 3) ML for Cyber-Agricultural Systems (October 2022). 4) Fundamentals of Deep Learning (November 2022) 5) Trending AI Techniques (April 2023) 6) Fifth International Workshop of Machine Learning for Cyber-Agricultural Systems (MLCAS 2023) workshop (July 2023) aimed to bring together academic and industrial researchers and practitioners in the fields of machine learning, data science and engineering, plant sciences and agriculture, to identify and discuss major technical challenges and recent results related to machine learning-based approaches. 7) AI for Agriculture and Food Systems, in conjunction with AAAI Conference and in collaboration with AgAID, AIFARMS, and AIFS. Intra-AIIRA Collaboration: We continue efforts to connect AIIRA faculty, researchers and graduate students across disciplines. For a second year, CMU team members visited Ames, IA to test their robotics in real world scenarios with ISU team members on site to provide context and valuable feedback. CMU made three visits to Iowa in the summer and fall of 2023 in 1) June 19-23 to field test the corn sensor insertion arm on the amiga robot, 2) July 24 - August 1 to field testing in soybeans, and 3) in mid-September to test the newly remodeled dexterous robot hand. Field Days: Engaging AIIRA's scientists, engineers and social scientists through well established, outward facing, field days and training workshops have been key to identifying the needs of our stakeholders and toward disseminating AI advances and impacts to agriculture. 1) The Workshop on Survey design and Experimental Methods in Applied and Agricultural Economics resulted in identifying recommendations for stakeholder engagement and policy design, 2) field demonstrations at the National Association of Plant Breeders Field Tour resulted in sharing research on sensors, root imaging and abiotic stress breeding, 3) Iowa Soybean Association field demonstrations showcasing research progress, and 4) collaborations with the HIPS and AG2PI offer opportunities for AIIRA to participate in planned field days Spring and Summer 2023. Social Media: With 356 Twitter followers, 137 LinkedIn followers and 56 Instagram followers, our outreach has grown over the past year. Posts shared include unique AIIRA content and sharing of AIIRA and AIIRA Institution related themes. Multi-organizational Synergies / Achievements. AIIRA strives to create community amongst our team members and partner institutions. AIIRA activity participates in AIVO efforts connecting with other AI Institutes on multiple fronts including sharing lessons learned with new AI Institutes. Knowledge Transfer. Our knowledge transfer and dissemination of results to research communities and communities of practice has been productive in our first year. While there have been no patents submitted or technology transfer, we have been productive in generating publications, presentations, events, activities, and other products Broadening Participation: AIIRA aims to apply best practices for mentoring and engaging women, Hispanics, and Native Americans in AI using evidence-based strategies across these underrepresented populations. (1) AIIRA, in collaboration with the Native Nations Institute at University of Arizona developed a Software Carpentry based learning material for ML/AI instruction with an Indigenous data focus. Progress on the Carpentries course development continues to move forward creating trainings and workshops and we will also develop a new Lesson Program. (2) The Indigenous Ag/Farmers Workshop occurred November 2022. This facilitated virtual event gathered diverse Indigenous agriculture practitioners, data science, and Indigenous Data Sovereignty scholars to discuss agriculture data priorities of Indigenous Peoples in the US. Designed and facilitated by an Indigenous led steering committee, representing 8 different tribal nations throughout the United States, participants a) identified practical uses of AI as: Monitoring, modeling, remote sensing, and to better understanding scale b) learned the application of AI and machine learning mean establishing culturally defined definitions of success and focus on center Indigenous values and c) participants left looking for more trainings, workshops, and live examples. (3) AIIRA's focus on Women in Ag and AI has grown through the continued efforts of our WIAA program. WIAA students have also 1) complete field demonstrations for undergraduate students on applications of UAVs and rovers in agricultural students, 2) discussed research, graduate education, and application of AI in agriculture with undergraduate agronomy students, 3) led a demonstration at the 4H Youth Leader Ag Innovator Experience on the insect app, UAV applications, sensors used in stress detection, and 3D/AR/VR field reconstruction to 9th through 12th graders in Iowa, 4) demonstrated the insect app, rover, and UAV to over 200 4th and 5th graders in Iowa, and 5) lead and activity at the WISE Go Further Spring 2023 event to engage and teach 8th grade students how rovers and AI are used in agriculture for insect detection and identification. AIIRA has also disseminated results through 39 publications, 43 presentations, 9 activities, 19 events, and 16 other products (listed in this report). What do you plan to do during the next reporting period to accomplish the goals?Research activities: We will continue to expand, and make progress on research activities initiated in Years 1-2. Our goal will be to wrap up a several projects by the end of Year 3, and initiate a few new projects. Selected new projects include (1) Integrate geometric and mechanics based constraints into 3D reconstruction, (2) Extend 3D geometry aware proximal sensing to hyperspectral modalities, (3) Enhance manipulation skills to incorporate domain knowledge, (4) Methods for aligning predictions with expert judgment Education and workforce development: Continue and expand on activities initiated in year 1-2, as well as start the following programmatic activities in Year 3: (1) An expanded suite of onboarding workshops including targeted education modules, (2) Regular virtual journal clubs for deep dive into key multidisciplinary papers, learning a common language, (3) Extension Showcases to Growers, (4) Start planning for an Applied AI minor at ISU. Broadening participation plans: Continue and expand on activities initiated in year 1-2: (1) Continue growing youth program with hands-on activities that explore intersection of AI and plant science in urban settings, (2) public workshops for bringing together various stakeholders and other AI Institutes, 3) explore strategic partnerships with other native nation communities, as well as minority serving institutions, 4) leverage the ExpandAI@NSF initiative to connect with MSI institutions

Impacts
What was accomplished under these goals? Goal 1: Build plant and field scale predictive models through foundational AI advances. Accomplishments in methodologies to build the digital twins: 1) We trained AI models for insect-pest detection to classify and detect agriculturally important insect species. We integrated the model with advanced features (out-of-distribution detection and conformal predictions) to enhance robustness. 2) We developed and deployed approaches for 3D reconstruction of plants. 3) We developed and are currently testing procedural models for representing 3D phenotypes. 4) We deployed and accessed zero-shot approaches (including vision-language models) for efficient phenotyping. 5) Developed foundational advances in mathematical methods, algorithms and neural solvers for solving PDEs with a focus on creating digital twins. Accomplishments in methodologies for feedback via AI-enabled robotics: 1) Developed a pipeline for field mobile manipulation integration and corn stalk sensor insertion. 2) Continued to develop dexterous robot hand for ag applications and a versatile spectroscopic probe infrastructure. 3) Developed several frugal, portable chambers for inducing heat, drought and flooding stress. 4) We deployed a suite of co-registered multi-modal sensors on ground- and aerial- vehicles. Goal 2: Deploy plant and field scale predictive models for breeding and crop production applications. 1) We have instituted multi-pronged research activities designing and evaluating a spectrum of prediction models. Continue advancing approaches to integrate data driven and crop model based ag predictors 2) We initiated a large set of data collection campaigns this growing season. 3) Started integrating physics and digital twins with preliminary experiments in hydroponics. 4) Developing novel phenotypic descriptors for 3D plant phenotyping. 5) Deployed several AI-driven phenotyping applications. 6) Deploying AI driven phenotyping and decision support to other crops. Goal 3: Understanding and resolving social barriers to, and AI innovations for adoption of the AI technology in the agricultural ecosystem 1) We are investigating various types of feedback (on adoption of AI) from targeted audiences, including immigrant farmers and organic farmers in and around Iowa, via interviews and small group focus groups. 2) Target Market, Customer Acquisition, Social Media, and Technology adoption among Immigrant, Rural, and Small farmers: Role of field Data Compilation and AI, 3) We are continuing our data collection efforts that combine translational research and behavioral economics to understand AI adoption in agriculture, 4) We leverage a large survey instrument to perform a demographic Analysis of Iowa Farmers and their perceptions towards data governance and benefits of precision ag adoption. 5) We also are making progress in understanding the (lack of) utility of NLP explanation benchmarks, along with other foundational advances in improving trustworthiness of AI systems. Goal 4: Create a diverse, AI-aware agricultural workforce and serve as a nexus for AI-in-Ag developments: Described in detail in next section.

Publications

  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: A. Nakkab, B. Feuer, C. Hegde, LiT-Tuned Models for Efficient Species Detection, AAAI AIAFS Workshop, February 2023. (Invited to Special Issue of Plant Phenomics, 2023.)
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Aditya Balu, Manoj R. Rajanna, Joel Khristy, Fei Xu, Adarsh Krishnamurthy, Ming-Chen Hsu; Direct immersogeometric fluid flow and heat transfer analysis of objects represented by point clouds, Computer Methods in Applied Mechanics and Engineering, 404(115742), 2023
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Anjana Deva Prasad, Aditya Balu, Harshil Shah, Soumik Sarkar, Chinmay Hegde, Adarsh Krishnamurthy; NURBS-Diff: A differentiable programming module for NURBS, Computer Aided Design, 146(103199), 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Anjana Deva Prasad, Anushrut Jignasu, Zaki Jubery, Soumik Sarkar, Baskar Ganapathysubramanian, Aditya Balu, Adarsh Krishnamurthy; Deep implicit surface reconstruction of 3D plant geometry from point cloud, AAAI workshop on AI for Agriculture and Food Systems (AIAFS), 2022
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Anushrut Jignasu, Ethan Herron, Talukder Zaki Jubery, James Afful, Aditya Balu, Baskar Ganapathysubramanian, Soumik Sarkar, Adarsh Krishnamurthy; Plant geometry reconstruction from field data using neural radiance fields, AAAI workshop on AI for Agriculture and Food Systems (AIAFS), 2023.
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Arjun Lakshmipathy, Nicole Feng, Yu Xi Lee, Moshe Mahler, and Nancy Pollard. Artist Tools for Intuitive Modeling of Hand-Object Interactions, ACM Transactions on Graphics. 2023
  • Type: Conference Papers and Presentations Status: Submitted Year Published: 2023 Citation: B. Feuer, A. Joshi, M. Cho, K. Jani, S. Chiranjeevi, C. Deng, A. Balu, A. K. Singh, S. Sarkar, N. Merchant, A. Singh, B. Ganapathysubramanian, C. Hegde, Zero-shot Insect Detection via Weak Language Supervision, AAAI AIAFS Workshop, February 2023. (Submitted to The Plant Phenome Journal, July 2023.)
  • Type: Journal Articles Status: Under Review Year Published: 2023 Citation: B. Feuer, A. Joshi, M. Pham, C. Hegde, Distributional Robustness on a Data Budget, under review, 2023
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Carley CN, MJ Zubrod, S Dutta, AK Singh. (2022). Using machine learning enabled phenotyping to characterize nodulation in three early vegetative stages in soybean. Crop Science, 00, 1 23. https://doi.org/10.1002/csc2.20861
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: SoftTouch: A Sensor-Placement Framework for Soft Robotic Hands, in IEEE-RAS 21st International Conference on Humanoid Robots (Humanoids), Okinawa, Japan. DOI: 10.1109/Humanoids53995.2022.10000138
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Chiteri, K.O., Chiranjeevi, S., Jubery, T.Z., Rairdin, A., Dutta, S., Ganapathysubramanian, B. and Singh, A., 2023. Dissecting the genetic architecture of leaf morphology traits in mungbean (Vigna radiata (L.) Wizcek) using genome?wide association study. The Plant Phenome Journal, 6(1), p.e20062.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Chiteri, K.O., Jubery, T.Z., Dutta, S., Ganapathysubramanian, B., Cannon, S. and Singh, A., 2022. Dissecting the Root Phenotypic and Genotypic Variability of the Iowa Mung Bean Diversity Panel. Frontiers in Plant Science, 12, p.808001.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: D. Malik, C. Igoe, Y. Li and A. Singh. Weighted Tallying Bandits: Overcoming Intractability via Repeated Exposure Optimality. International Conference on Machine Learning, ICML'23
  • Type: Other Status: Published Year Published: 2023 Citation: Soft Robotic End-Effectors in the Wild: A Case Study of a Soft Manipulator for Green Bell Pepper Harvesting. In AI for Agriculture and Food Systems (AIAFS) Workshop, AAAI Conference on Artificial Intelligence 2023, Washington, DC
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Dong, D., Nagasubramanian, K., Wang, R., Frei, U.K., Jubery, T.Z., L�bberstedt, T. and Ganapathysubramanian, B., 2023. Self-supervised maize kernel classification and segmentation for embryo identification. Frontiers in Plant Science, 14, p.1108355.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Elango D, K Rajendran, L Van der Laan, S Sebastiar, J Raigne, NA Thaiparambil, N El Haddad, B Raja, W Wang, A Ferela, KO Chiteri, M Thudi, RK Varshney, S Chopra, A Singh, AK Singh (2022). Raffinose Family Oligosaccharides: Friend or Foe for Human and Plant Health?. Frontiers in plant science, 13, 829118. https://doi.org/10.3389/fpls.2022.829118
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Herr, A. W., Adak, A., Carroll, M. E., Elango, D., Kar, S., Li, C., Jones, S. E., Carter, A. H., Murray, S. C., Paterson, A., Sankaran, S., Singh, A., & Singh, A. K. (2023). Unoccupied aerial systems imagery for phenotyping in cotton, maize, soybean, and wheat breeding. Crop Science, 63, 1722 1749. https://doi.org/10.1002/csc2.21028
  • Type: Journal Articles Status: Other Year Published: 2023 Citation: J. Hsia, D. Purthi, Z. Lipton and A. Singh. Goodharts Law Applies to NLPs Explanation Benchmarks. To Be Submitted.
  • Type: Journal Articles Status: Under Review Year Published: 2023 Citation: Jiale Feng, Mojdeh Saadati, Talukder Jubery, Anushrut Jignasu, Aditya Balu, Yawei Li, Lakshmi Attigala, Patrick Schnable, Soumik Sarkar, Baskar Ganapathysubramanian, Adarsh Krishnamurthy, 3D Reconstruction of Plants Using Probabilistic Voxel Carving, Computers and Electronics in Agriculture, Under Review, 2023.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Khanwale, Makrand A., Kumar Saurabh, Milinda Fernando, Victor M. Calo, Hari Sundar, James A. Rossmanith, and Baskar Ganapathysubramanian. "A fully-coupled framework for solving Cahn-Hilliard Navier-Stokes equations: Second-order, energy-stable numerical methods on adaptive octree based meshes." Computer Physics Communications 280 (2022): 108501.
  • Type: Journal Articles Status: Accepted Year Published: 2023 Citation: Yanbin Chang, Jeremy Latham, Mark Licht, and Lizhi Wang, "A data-driven crop model for maize yield prediction," to appear in Communications Biology, 2023.
  • Type: Conference Papers and Presentations Status: Awaiting Publication Year Published: 2023 Citation: Zheng Ni, Saba Moeinizade, Aaron Kusmec, Guiping Hu, Lizhi Wang, and Patrick S. Schnable, "New insights into trait introgression with the look-ahead intercrossing strategy," to appear in G3, 2023.
  • Type: Journal Articles Status: Accepted Year Published: 2023 Citation: Isaiah Huber, Lizhi Wang, Jerry L. Hatfield, Mark Hanna, and Sotirios V. Archontoulis, "Modeling days suitable for fieldwork using machine learning, process-based, and rule-based models," to appear in Agricultural Systems, 2023.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Zahra Khalilzadeh and Lizhi Wang, "Corn planting and harvest scheduling under storage capacity and growing degree units uncertainty," to appear in Scientific Reports, 2022.
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Kar, S., Nagasubramanian, K., Elango, D., Carroll, M. E., Abel, C. A., Nair, A., Mueller, D. S., ONeal, M. E., Singh, A. K., Sarkar, S., Ganapathysubramanian, B., & Singh, A. (2023). Self-supervised learning improves classification of agriculturally important insect pests in plants. The Plant Phenome Journal, 120. https://doi.org/10.1002/ppj2.20079
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Khanwale, M.A., Saurabh, K., Ishii, M., Sundar, H., Rossmanith, J.A. and Ganapathysubramanian, B., 2023. A projection-based, semi-implicit time-stepping approach for the Cahn-Hilliard Navier-Stokes equations on adaptive octree meshes. Journal of Computational Physics, 475, p.111874.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Saeed Khaki, Hieu Pham, Zahra Khalilzadeh, Arezoo Masoud, Nima Safaei, Ye Han, Wade Kent, Lizhi Wang, "High-throughput image-based plant stand count estimation using convolutional neural networks," to appear in PLOS ONE, 2022.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Zerui Zhang, Lizhi Wang. "A look-ahead approach to maximizing present value of genetic gains in genomic selection," G3, vol. 12(8), 2022.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Saba Moeinizade, Guiping Hu, and Lizhi Wang, "A reinforcement learning approach to resource allocation in genomic selection," to appear in Intelligent Systems with Applications, 2022.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Fatemeh Amini, Guiping Hu, Lizhi Wang, and Ruoyu Wu, "The L-shaped selection algorithm for multi-trait genomic selection," to appear in Genetics, 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: M. Cho, et al, Sphynx: A Deep Neural Network Design for Private Inference, IEEE S&P, October 2023
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Nagasubramanian K, AK Singh, A Singh, S Sarkar, B Ganapathysubramanian. (2022). Plant phenotyping with limited annotation: Doing more with less. The Plant Phenome Journal, 5, e20051. https://doi.org/10.1002/ppj2.20051
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Rairdin A, F Fotouhi, J Zhang, DS Mueller, B Ganapathysubramanian, AK Singh, S Dutta, S Sarkar, and A Singh. (2022). Deep learning-based phenotyping for genome wide association studies of sudden death syndrome in soybean. Frontiers in plant science, 13, 966244. https://doi.org/10.3389/fpls.2022.966244
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2023 Citation: Siva Kailas, Wenhao Luo, and Katia Sycara. Multi-robot adaptive sampling for supervised spatiotemporal forecasting. In Progress in Artificial Intelligence: 22nd EPIA Conference on Artificial Intelligence, EPIA 2023. Springer, 2023.
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Webster RW, M McCaghey, B Mueller, C Groves, FM Mathew, A Singh, M Kabbage, DL Smith. (2023). Development of Glycine max Germplasm Highly Resistant to Sclerotinia sclerotiorum. PhytoFrontiers. https://doi.org/10.1094/PHYTOFR-01-23-0009-R
  • Type: Journal Articles Status: Accepted Year Published: 2023 Citation: High-Throughput Field Plant Phenotyping: A Self-Supervised Sequential CNN Method to Segment Overlapping Plants
  • Type: Other Status: Awaiting Publication Year Published: 2023 Citation: A Systematic Error Cleaning and Correction Pipeline for Field High-Throughput Phenotyping
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Young TJ, TZ Jubery, CN Carley, M Carroll, S Sarkar, AK Singh, A Singh, B Ganapathysubramanian. (2023). "Canopy fingerprints" for characterizing three-dimensional point cloud data of soybean canopies. Frontiers in plant science, 14, 1141153. https://doi.org/10.3389/fpls.2023.1141153


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

Outputs
Target Audience:AIIRA reaches a broad and diverse audience of USDA/NIFA stakeholders through our research and outreach activities. • K-12 students (middle and high school) including students of color and students from economically disadvantaged backgrounds • Graduate students and postdoctoral fellows • Undergraduate and community college students including women, native american and hispanic populations • Farmers and commodity groups including Indigenous Farmers • Extension and teaching professionals, Ag professionals • USDA ARS scientists • Industry Changes/Problems:Management Team: C. Lawrence-Dill (ISU) has transitioned from a management team member to an ex-officio role due to her promotion into a leadership position within Iowa State University and appointment to the FFAR Board of Directors. Now serving as the Associate Dean of Research for the College of Agriculture and Life Sciences, Lawrence-Dill has valuable experience and expertise, and she continues her contributions to AIIRA with her graduate students and their projects. AIIRA's management structure includes a succession plan that lays out the process for this transition. Her role as co-lead of the Education and Workforce Development Thrust has been transitioned to A. Johri (GMU). Other personnel: 1) J. Schnable (UNL) has taken a leave of absence to join Google X, the Moonshot Factory, for four months. We anticipate that his experience at X will be highly beneficial to AIIRA in terms of identifying opportunities and synergies. Ravi Mural will temporarily step in for J. Schnable for Fall 2022. 2) Peter Kyveryga has taken a new position outside of the Iowa Soybean Association and his role is temporarily being filled by Aaron Presholt until Kyveryga's position is filled. Kyveryga has joined John Deere and will be a useful external collaborator in future. Presholt has been tangentially involved in AIIRA broader impact efforts and the transition to fulfilling Kyveryga's responsibilities has progressed smoothly. Hiring: Hiring students and staff was a challenge due to the COVID pandemic. Visa processes for international students/staff resulted in long wait-times. We anticipate this to be a temporary issue, with rapidly shortening wait-times suggesting this is no longer an issue in year 2 and beyond. Budget: 1) Due to our Oct 1 start date, along with timing of standard administrative processes involved in subcontract issuance, we underspent our budget for Y1. We anticipate catching up to planned budget burn-rates in the next reporting cycles. 2) No-cost extension is needed to maintain graduate research assistantships from year 1 into year 2, and potentially for future years. What opportunities for training and professional development has the project provided?AIIRA's education and workforce development opportunities aim to develop talent and skills for a highly competent next generation AI workforce which includes activities for graduate and undergraduate students, and continual learners of all ages. A diverse team from across AIIRA's collaborative institutions aim to leverage existing partnerships to create new learning opportunities and expand existing programs aimed at advancing knowledge and education in AI in Ag. University and college: AIIRA personnel directly mentored 3 postdocs, 53 graduate students, and 11 undergraduates across 8 institutions. Mentoring consisted of both formal course work, training and research assistantship, as well as informal networking across institutions. We briefly detail selected activities below, with additional details in the full report to the cognizant program manager. Undergrad/grad education: 1) We made progress planning for undergraduate programming in AI in Ag. The AI for Robotics course developed at CMU forms the basis for a proposed four course AI in Ag sequence that will be expanded at other AIIRA institutions. 2) GMU has prepared a case study course on ethics of use in AI for Ag that is deploying this fall. 3) Additionally, AIIRA faculty served as guest lecturers for CMU's new Robotics and AI for Ag course in Spring 2022. 4) ISU has developed an undergraduate minor in engineering called Cyber-Physical Systems with modules planned for AI in Ag. 5) We are planning for Ag-AI content in the recently offered Data Science minor at ISU, focused on undergrads from Ag and Life Sciences. 5) New course on "foundations in deep learning" at NYU. One of the first course of its kind nationwide. Workforce development via Ag-AI related training opportunities: 1) Application development exploring various AI and software tools with three honors freshmen, one of which built a prototype for their summer research. 2) AI, digital transformation and digital agriculture research involving one honors undergraduate and an MBA graduate student, 3) Student mentoring on entrepreneurial marketing projects in digital agriculture, agricultural and applied economics as it relates to ethical and social barriers of AI in Ag. 4) Four graduate and two undergraduate students worked on different aspects of AI, AI Ethics and AI ethics and agriculture related projects resulting in created case studies for instruction implemented in an undergraduate course in Fall 2022. 5) The development of ML algorithms by two undergraduate students, who took into account interaction with humans and the capabilities of human feedback. 6) Exposed a sustainable agriculture & sociology PhD student to the social dimensions of AI in Ag literature and research. 7) Several engineering undergrads exposed to AI-in-Ag research opportunities. 8) Engineering capstone design teams. Workforce development via connecting with stakeholders: AIIRA completed activities to generate a new AI-in-Ag workforce. Efforts were focused on identifying means to connect with farmers and growers to demonstrate AI driven tools and technologies as well as means to showcase AI driven technologies to the broader agricultural community. 1) Workshop on AI, Ag, Ethical and Responsible Development which attracted participants from the industry, postdocs and students from other AI & AG institutes, and other participants from academia and non-profits. 2) AIIRA presence at the 2022 Farm Progress Show in Boone, IA. 3) The Data Science for the Public Good, Wholesale Local Food Benchmarking Project sponsored by AIIRA and managed by a graduate fellow and three undergraduate interns. The DPSG students learned and practiced data discovery and practical applications of AI in Ag. Workforce development via community building activities: Organized several workshops to disseminate research and education advances to various stakeholders. 1) Tutorials on using HPC with machine learning at the CVPR conference and the SuperComputing conference, 2) Workshop on Scientific Machine Learning at ISU, 3) International Workshop on Machine Learning for Cyber Ag systems (MLCAS) in 2021 (virtual), and scheduled for 2022 (Ames, IA), 4) AAAI workshop on AI for Ag and Food Systems (virtual), 5) Presented AIIRA vision and collaborative opportunities at various venues (USDA SAS PI meeting, AG2PI Field days, Farm Progress Show, National Association for Plant Breeders (NAPB) meetings, NSF ERC IoT4Ag seminar series, State of Iowa Legislators, IA Board of Regents, US & Mexico Secretaries of Ag). Pre-college education: 1) AIIRA's pre-college education efforts focused on leveraging existing partnerships at CMU with a goal to develop a new curriculum in Pittsburgh, PA through the Urban Robotony workshop for middle schoolers. As a signature AIIRA effort, the program is a hands-on program targeting middle school students that combines AI, robotics, and plant science. The first year of the program was dedicated to building partnerships, selecting a hardware platform and developing curriculum for the pilot, which was offered in August 2022. The pilot took the form of a 1-day, 3-hour workshop offered at the Drew Matheison Greenhouse in Pittsburgh and included an introduction to plant science, AI technologies for Ag, and a hands-on activity using the FarmBot platform. There were 45 students in attendance, including 9-12th grader students and students in Drew Matheison's Horticulture Technology workforce development program. 2) ISU's summer internships resulted in five high school students learning machine learning coding as it relates to field corn and common agricultural insects. AIIRA is sponsoring a unique informal learning opportunity to engage with undergraduate students outside of the AI and Agriculture frames. A. Johri (GMU) and a graduate student are planning a writing competition to bring a diverse audience to generate ideas for AI solving agricultural problems. Beginning in August 2022, undergraduates across the United States are invited to submit writings that imagine the world in the year 2050 and explore the relationship between humans, technology, and their food. The goal is for this opportunity to expand in the future to attract a broader audience to AI-in-Ag and generate further creative solutions to today's problems. How have the results been disseminated to communities of interest?AIIRA results are disseminated through collaboration and knowledge transfer workshops and training, sharing of benchmark datasets, and participation in field days to engage directly with stakeholders. Collaboration and knowledge transfer: AIIRA has worked towards creating a learning community to support a collaborative research and training environment across the transdisciplinary boundaries of team members. AIIRA is administratively housed within the Center for Translational AI (TrAC) at ISU. With the goal of providing the means to support collaboration and sharing of knowledge, and using team science based approaches, the following workshops and trainings have been established or shared: AIIRA sponsored workshops and training: AIIRA sponsored, developed and implemented several workshops and training opportunities to AIIRA stakeholders. 1) TrAC Journal Club included AIIRA graduate students who participated in 7 sessions offered in Spring 2022. 2) TrAC Seminar Series (offered virtually) sponsored by AIIRA presented to a wide audience across disciplines including a) Oliver Kroemer of Carnegie Mellon University (CMU) Robotics Institute (Jan 28, 2022), b) Mark Ryan of Wageningen Economic Research (Feb 25, 2022), c) Volkan Isler of University of Minnesota/Samsung Research (April 15, 2022), d) Brian Bailey of University California Davis, presenting on Helios (May 5, 2022), 3) Fundamentals of Deep Learning with Multi-GPUs - in collaboration with NVIDIA Deep Learning Institute (July 29). 4) Third International Workshop of Machine Learning for Cyber-Agricultural Systems (MLCAS2021) workshop (November 2 - 4, 2021) aimed to bring together academic and industrial researchers and practitioners in the fields of machine learning, data science and engineering, plant sciences and agriculture, in the collaborative effort of identifying and discussing major technical challenges and recent results related to machine learning-based approaches. 5) Intro to Cloud Based Deep Learning, in collaboration with CyVerse, offered an introduction to concepts in cloud, containerization, and GPU computing with Python. 6) APSIM Informational Webinar presented by S. Archontoulis on the Agricultural Production Systems sIMulator (APSIM) and how the cropping systems model is capable of simulating the growth and development of many crops on a daily basis. 7) FarmBot Workshop (Urban Robotany) by George Kantor at the Carnegie Mellon University (CMU) Robotics Institute offering an introduction to plant science, AI technologies for Ag, and a hands-on activity using the FarmBot platform (August 18, 2022). Workshops and training promoted to AIIRA: The following opportunities were shared broadly to further developing cross disciplinary, cross institutional relationships and to further the mission of AIIRA. 1) Plant Phenomics Phridays - Held on Fridays at ISU, Speakers in this series describe novel solution approaches to plant science problems from different perspectives spanning the plant sciences, statistics, data sciences and engineering. 2) International Conference on Digital Technologies for Sustainable Crop Production (DIGICROP), March 28-30, 2022 - AIIRA team members participated in the opportunity to develop, propose, use or evaluate new digital technologies across the intersection of engineering, robotics, crops sciences, computer science, agricultural sciences, economics and phenotyping. 3) Virtual Collaborative Workshop for Early-Career Academics via the KY EPSCoR office, May 16, 2022 - opportunity shared to learn more about leadership style and how to build programs to expand the impact of research. 4) ICRA 2022 Workshop on Innovation in Forestry Robotics: Research and Industry Adoption, May 23, 2022 - addressing recent advances in robotics research and development for forestry applications. 5) AI FARMS - AI Foundry for Agricultural Applications Summer School - July 11-16th AIIRA graduate students and postdocs were encouraged to apply to increase their ability to engage in conversations and generate new ideas for AI applications to solve agricultural issues with AI. 5) CyVerse - Webinar: Zero to Web App: Rapid Customized Web Interfaces for Your ML Applications with Streamlit, August 26, 2022, Virtual, designed to learn how to use Streamlit, a low-code solution that allows you to quickly build web applications using Python scripts and How you can incorporate machine learning into your analysis workflow. A significant effort has been made to connect AIIRA faculty, researchers and graduate students across disciplines. Bringing 12 CMU team members from the Robotics Institute to Ames, IA for two weeks generated opportunities for the team to more easily understand the context in which they work. Having the robotics team test robots and drones in the field was highly beneficial to the research and knowledge sharing. Field Days: Engaging AIIRA's scientists, engineers and social scientists through well established, outward facing, field days and training workshops are key to identifying the needs of our stakeholders and toward disseminating AI advances and impacts to agriculture. 1) Iowa Soybean Association Tours - leveraged through ISA and Iowa Soybean Research Center, AIIRA faculty and students attended a field tour of Kemin Industries, Beck's Hybrids, and Salin247 to receive an overview of their research and test plots. 2) Farmbot Workshop [CMU] - Drew Matheison Center greenhouse, August 16 - 18, 2022, 3) 2022 Farm Progress Show, Boone IA, August 30 - Sept 1, 2022. Broadening Participation: AIIRA aims to apply best practices for mentoring and engaging women, Hispanics, and Native Americans in AI using evidence-based strategies across these underrepresented populations. (1) AIIRA, in collaboration with the Native Nations Institute at University of Arizona is developing a Software Carpentry based learning material for ML/AI instruction with an Indigenous data focus covering topics including Tribal Institutional Review Boards, The CARE Principles for Indigenous Data Governance, framework/s for data sharing, cultural autonomy, and the importance of consent in data. The global platform aligns with CELSI (Cultural, Ethical, Legal, and Social Implications) format to allow for content that is broader and outside of the scope of Indigenous topics. (2) In addition to the lesson program and courses, we will also have team members complete the Carpentiers Instructor Training and we will write a scholarly manuscript detailing the co-production of the new lesson program. (3) The Indigenous Ag/Farmers Workshop is in the planning stages for November 2022. The one-day facilitated, interactive event will be guided by a 4-5 member, scholar/practitioner Indigenous Steering Committee to structure the workshop to identify issues, opportunities and strategies for meeting Indigenous agricultural data needs. Workshop attendees will consist of industry leaders and Indigenous community members. (4) AIIRA's focus on Ag-Women-in AI has been highlighted in a recent video "Bots in Beans: Ag Women using AI Tools (Robots and UAVs in Ag)" featuring graduate Student Ashlyn Reardon (ISU) (https://youtu.be/PLfgCVFCeCY). (5) AIIRA has generated 270 organic followers across 3 platforms (Twitter, LinkedIn, and Instagram) since their creation in February 2022. Posts shared include unique AIIRA content and sharing of AIIRA and AIIRA Institution related themes. AIIRA has also disseminated results through 24 publications, 42presentations, 15 invited talks, 4 activities, 14 events, and 2 other products (listed in this report). What do you plan to do during the next reporting period to accomplish the goals?Research activities: We will continue to expand, and make progress on research activities initiated in Year 1. We will start the following additional activities in Year 2: (1) Integrate geometric and mechanics based constraints into 3D reconstruction, (2) Extend 3D geometry aware proximal sensing to hyperspectral modalities, (3) Design and deploy strategies for multi-modal data assimilation. (4) Enhance manipulation skills to incorporate domain knowledge, (5) Scale up to fieldable intelligent systems, (5) Enhance the capability of the production Ag decision support system, (6) Design and implement nationwide online surveys with a wide spectrum of farmers to elicit potential motivators and barriers for adopting new technologies with AI components. (7) Methods for aligning predictions with expert judgment, (8) Methods for robustification, and secure training of AI-enabled models. Education and workforce development: Continue and expand on activities initiated in year 1, as well as start the following programmatic activities in Year 2: (1) On boarding workshops including targeted education modules, (2) Weekly Seminars to network and learn the state-of-art landscape, (3) Weekly virtual journal clubs for deep dive into key multidisciplinary papers, learning a common language, (4) data science and other citizen science competitions for bringing together AI and Ag experts to energize the future workforce, (5) Ag-AI Graduate Minor Development at Iowa State, (8) Extension Showcases to Growers. Broadening participation plans: Continue and expand on activities initiated in year 1, as well as start the following programmatic activities in Year 2: (1) Establish youth program with hands-on activities that explore intersection of AI and plant science in urban settings, (2) public workshops for bringing together various stakeholders and other AI Institutes, 3) explore strategic partnerships with other native nation communities, as well as minority serving institutions.

Impacts
What was accomplished under these goals? Overall Impact Statement: We have made good progress in Year 1 towards accomplishing project objectives in each of thrusts. We have made foundational and translational advances impacting breeding and production agriculture; Our broadening participation, education and workforce development, and collaboration and knowledge transfer activities have already benefited a wide spectrum of stakeholders. Goal 1: Build plant and field scale predictive models through foundational AI advances. Accomplishments in sensor design and field deployment to feed the digital twins: 1) We developed a wearable, minimally invasive biosensor for quasi-continuous monitoring of salicylic acid in plant tissues. 2) We developed a wearable plant sensor that can monitor volatile organic compounds emission (specifically, methanol) directly on the leaf of a plant under field conditions with low cost and easy installation. 3) We redesigned the Episcan color sensor to capture direct and indirect light transport to image the surface and subsurface structures. Accomplishments in methodologies to build the digital twins: 1) We trained AI models for insect pest detection to classify agriculturally important insect-pesta. We also deployed this model into a user-friendly app for maximizing adoption and democratized use. 2) We developed and deployed two approaches for 3D reconstruction of plants: a) designed probabilistic approaches to 3D reconstruction from 2D multi-view images. b) reconstructed watertight STL surfaces of plants from 3D point cloud data. 3) We started development on detailed simulations of air flow, temperature and CO2 distributions at the plant scale by incorporating realistic 3D structures into fluid dynamics (CFD) simulations. Accomplishments in methodologies for feedback via AI-enabled robotics: 1) We designed and performed preliminary testing of a robotic end-effector for inserting nitrate sensors into corn stalks. 2) We developed and are currently testing methods for identifying parts of plants (e.g., leaves vs. stem) from vibrotactile feedback. 3) We deployed a suite of co-registered multi-modal sensors on ground- and aerial- vehicles. 4) We designed and performed preliminary testing of a soft robotic arm combined with a cutting tool for an initial attempt to harvest pepper plants in the field. Goal 2: Deploy plant and field scale predictive models for breeding and crop production applications. ML based vs process based vs hybrid crop models: We have instituted multi-pronged research activities designing and evaluating a spectrum of prediction models: 1) We made progress towards a new software system to run the APSIM crop model in parallel. The new software feeds the APSIM crop model with weather data, management data and soil data for every grid cell. Once complete, we will be able to run APSIM across scales, from subfield to field and to regional scale. 2) We built deep learning, as well as statistical models to extract phenotypic information from various field experiments (maize, soybean). These data were used to train yield prediction ML algorithms. We evaluated the ability to predict not only yield, but also other traits such as protein, and oil concentrations. 3) We built a hybrid model for crop yield prediction that combines process based crop models with data driven machine learning algorithms. As a follow up of this activity, we are starting to validate this model in hydroponic experiments. Optimizing Breeding Strategies: 1) We designed new algorithms for trait introgression, which challenges the widely used backcrossing breeding scheme with a more general and more effective intercrossing strategy. 2) We designed a breeding simulator and used it to organize a class competition. We plan to open source the simulator in three popular programming languages: Matlab, Python, and Julia. Farm management decision support tools: 1) Developed a simulator for a virtual farm with potential biotic stresses and stress mitigation strategies based on available knowledge in the literature and crop models. Developed an ML based decision support tool for optimizing management strategies using this simulator. 2) We considered an abstraction of the farming decision problem as an AI-enabled sequential decision-making-process. We leverage the specific memory-bounded structure and provide an algorithm along with its complete policy regret. We couple this result with a lower bound on the complete policy regret. Data collection campaigns: We have initiated a large set of data collection campaigns this growing season. 1) Microbe assisted selection for identifying heat tolerant genes and phenes in soybean. We conducted a factorial experiment with four soybean genotypes, with and without microbes, and heat versus optimum temperature. Experiments were performed in the Enviratron (a controlled environment facility at ISU) and we collected plant tissues for DNA, RNA, root exudate, and anatomical studies. 2) Heat stress tolerance screening to identify candidate genes using a large panel of soybean accessions. We screened 450 unique soybean genotypes for their tolerance to heat stress. 3) Development of field deployable greenhouse for heat tolerance screening. We partnered with two teams in a senior capstone class in engineering to develop portable and autonomous greenhouses for use in the field. This is to allow for simulation of heat stress conditions out in the field. 4) We collected temporally and spatially tagged multimodal data (drones, ground measurements, and high resolution satellite) across the geographically distributed, highly instrumented tests sites available through the USDA-NIFA High-Intensity Phenotyping Sites (HIPS) project, which is evaluating hundreds of maize inbreds and hybrids. Goal 3: Understanding and resolving social barriers to, and AI innovations for adoption of the AI technology in the agricultural ecosystem Social science activities: 1) Fielded a survey in 2022 among Iowa farmers which included a module on precision agriculture technology adoption. Conducted extensive original data analysis of survey data containing measures of precision agriculture adoption among Iowa farmers that was collected in 2017. 2) Built a software prototype to convert voice transcripts of the operational tasks from vegetable produce farner into a knowledge graph using AI and NLP. 3) Completed qualitative data analysis for a study on the digital agriculture ecosystem. 4) Analysis of data on digital agriculture technology and networking collected via in-person surveys with commodity farmers in Iowa in 2021 and 2022. 5) Creating marketing knowledge on how AI impacts relational dynamics amongst digital agriculture stakeholders and consumer behavior. AI activities: 1) Two new frameworks for designing AI models that respect privacy of stakeholders in the data processing pipeline, thereby enhancing end user trust. This work represents a significant improvement over the current state of the art. 2) Formulated and deployed out-of-distribution algorithms for insect-pest classification. This allows the model to abstain from prediction in a robust manner, further enabling end user trust. 3) We developed a new approach for interrogating and enhancing robustness of a particular family of neural networks called vision transformers (ViTs) which are the current state of the art. 4) We are investigating various types of feedback that experts can provide and their value for an AI system. Goal 4: Create a diverse, AI-aware agricultural workforce and serve as a nexus for AI-in-Ag developments: Described in detail in next section.

Publications

  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Leikas, J., Johri, A., Latvanen, M., Wessberg, N. & Hahto A. (2022). Governing Ethical AI Transformation: Case Aurora AI. Frontiers in Artificial Intelligence: AI for Human Learning and Behavior Change.
  • Type: Other Status: Accepted Year Published: 2021 Citation: Cho, M., Nagasubramanian, K., Singh, A.K., Singh, A., Ganapathysubramanian, B., Sarkar, S., Hegde, C. (2021) Privacy-Preserving Deep Models for Plant Stress Phenotyping
  • Type: Other Status: Accepted Year Published: 2021 Citation: Kar, S., Nagasubramanian, K., Elango, D., Nair, A., Mueller, D.S., ONeal, M.E., Singh, A.K., Sarkar, S., Ganapathysubramanian, B., and Singh, A. (2021) Self-Supervised Learning Improves Agricultural Pest Classification
  • Type: Other Status: Accepted Year Published: 2021 Citation: Chiranjeevi, S., Young, T., Jubery, T.Z., Nagasubramanian, K., Sarkar, S., Singh, A.K., Singh, A., and Ganapathysubramanian, B. (2021). Exploring the use of 3D point cloud data for improved plant stress rating
  • Type: Other Status: Accepted Year Published: 2021 Citation: Nagasubramanian, K., Singh, A.K., Singh, A., Sarkar, S., and Ganapathysubramanian, B. (2021). Plant Phenotyping with Limited Annotation: Doing More with Less
  • Type: Conference Papers and Presentations Status: Awaiting Publication Year Published: 2022 Citation: Bauer, Dominik, Cornelia Bauer, Arjun Lakshmipathy, Roberto Shu, and Nancy S. Pollard. "Towards Very Low-Cost Iterative Prototyping for Fully Printable Dexterous Soft Robotic Hands." International Conference on Soft Robotics (ROBOSOFT) 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Fully Printable Low-Cost Dexterous Soft Robotic Manipulators for Agriculture, D Bauer, C Bauer, A Lakshmipathy, N Pollard - AI for Agriculture and Food Systems, 2021
  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2022 Citation: Lakshmipathy, Arjun, Dominik Bauer, Cornelia Bauer, and Nancy S. Pollard. "Contact Transfer: A Direct, User-Driven Method for Human to Robot Transfer of Grasps and Manipulations." IEEE International Conference on Robotics and Automation (2022).
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Johri, A. (2022). Augmented Sociomateriality: Implications of Artificial Intelligence for the Field of Learning Technology. Research in Learning Technology.
  • Type: Other Status: Published Year Published: 2022 Citation: Johri, A. & Hingle, A. (2022). Learning to Link Micro, Meso, and Macro Ethical Concerns Through Role-Play Discussions. Proceedings of FIE 2022.
  • Type: Other Status: Published Year Published: 2022 Citation: Hingle, A. & Johri, A. (2022). Assessing Engineering Students Representation and Identification of Ethical Dilemmas through Concept Maps and Role-Plays. Proceedings of ASEE 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Dhruv Malik; Yuanzhi Li; Aarti Singh. Complete Policy Regret Bounds for Tallying Bandits. Proceedings of Conference on Learning Theory (COLT), 2022.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Ibraham, H., Satyanarayana, M., Schnable, P., & Dong, L. (2022). Wearable plant sensor for In Situ monitoring of volatile organic compound emissions from crops. American Chemical Society https://doi.org/10.1021/acssensors.2c00834
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Ansarifar J, Wang L, Archontoulis SV, 2021. An interaction regression model for crop yield prediction. Nature Scientific Reports 11, 17754.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Shahhosseini M, Hu G, Khaki S, Archontoulis SV, 2021. Corn yield prediction with ensemble CNN-DNN. Frontiers Plant Science 12, 709008. doi: 10.3389/fpls.2021.709008
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Akhavizadegan F, Wang L, Huber I, Archontoulis S, 2021. A Time-Dependent Parameter Estimation Framework for Crop Modeling. Nature Scientific Reports 11:11437
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Standardized genome-wide function prediction enables comparative functional genomics: a new application area for Gene Ontologies in plants L Fattel, D Psaroudakis, CF Yanarella, KO Chiteri, HA Dostalik, P Joshi, ... GigaScience 11 https://doi.org/10.1093/gigascience/giac023
  • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Jayashankar, P, Johnston, W., and Nilakanta, S., How do agricultural stakeholders perform institutional work through AI? Macro-marketing Conference, 2021.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2021 Citation: Jayashankar, P, Johnston, W., and Nilakanta, S., Market-shaping through B2B value chain reconfiguration  a study of digital twins, CBIM International Conference, 2021.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Hegde, C., Cho, Ghodsi, Reagen, Garg. Sphynx: ReLU efficient network design for Private Inference, IEEE S&P, 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Hegde, C., Cho, Joshi, Reagen, Garg. Selective Network Linearization for Private Inference, ICML 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Hegde, C., Cho, Singh, Singh, Sarkar, Koushik N., Ganapathysubramanian. Privacy-preserving Deep Models for Plant-Stress Phenotyping, AAAI Workshop, AIAFS 2022.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Hegde, C., Joshi, Jagatap. Adversarial Token Attacks on Vision Transformers, CVPR 2022 Workshop on Transformers for Vision.
  • Type: Conference Papers and Presentations Status: Published Year Published: 2022 Citation: Singh, A., Saadit, Ganapathysubramanian. OOD Algorithms for Robust Insect-Pest Classification
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Zhou, Y., Kusmec, A., Mirnezami, S.V., Attigala, L., Srinivasan, S., Jubery, T.Z., Schnable, J.C., Salas-Fernandez, M.G., Ganapathysubramanian, B. and Schnable, P.S., 2021. Identification and utilization of genetic determinants of trait measurement errors in image-based, high-throughput phenotyping. The Plant Cell, 33(8), pp.2562-2582.
  • Type: Journal Articles Status: Published Year Published: 2022 Citation: Chiteri, K.O., Jubery, T.Z., Dutta, S., Ganapathysubramanian, B., Cannon, S. and Singh, A., 2022. Dissecting the root phenotypic and genotypic variability of the Iowa mung bean diversity panel. Frontiers in plant science, 12, p.808001.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Jubery, T.Z., Carley, C.N., Singh, A., Sarkar, S., Ganapathysubramanian, B. and Singh, A.K., 2021. Using machine learning to develop a fully automated soybean nodule acquisition pipeline (snap). Plant Phenomics, 2021.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Riera, L.G., Carroll, M.E., Zhang, Z., Shook, J.M., Ghosal, S., Gao, T., Singh, A., Bhattacharya, S., Ganapathysubramanian, B., Singh, A.K. and Sarkar, S., 2021. Deep multiview image fusion for soybean yield estimation in breeding applications. Plant Phenomics, 2021.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Shook, J., Gangopadhyay, T., Wu, L., Ganapathysubramanian, B., Sarkar, S. and Singh, A.K., 2021. Crop yield prediction integrating genotype and weather variables using deep learning. Plos one, 16(6), p.e0252402.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Guo, W., Carroll, M.E., Singh, A., Swetnam, T.L., Merchant, N., Sarkar, S., Singh, A.K. and Ganapathysubramanian, B., 2021. UAS-based plant phenotyping for research and breeding applications. Plant Phenomics, 2021.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Kusmec, A., Zheng, Z., Archontoulis, S., Ganapathysubramanian, B., Hu, G., Wang, L., Yu, J. and Schnable, P.S., 2021. Interdisciplinary strategies to enable data-driven plant breeding in a changing climate. One Earth, 4(3), pp.372-383.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Mirnezami SV, Srinivasan S, Zhou Y, Schnable PS, Ganapathysubramanian B. Detection of the progression of anthesis in field-grown maize tassels: A case study. Plant Phenomics. 2021 Mar 3;2021.
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Nagasubramanian, K., Jubery, T., Fotouhi Ardakani, F., Mirnezami, S.V., Singh, A.K., Singh, A., Sarkar, S. and Ganapathysubramanian, B., 2021. How useful is active learning for image?based plant phenotyping?. The Plant Phenome Journal, 4(1), p.e20020.