ADVANCEA: ADVANCING CONTROLLED ENVIRONMENT AGRICULTURE THROUGH DATA-DRIVEN DECISION MAKING AND WORKFORCE DEVELOPMENT

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: ACTIVE
• Funding Source: OTHER GRANTS
• Reporting Frequency: Annual
• Accession No.: 1029246
• Grant No.: 2022-51181-38324
• Cumulative Award Amt.: $3,773,807.00
• Proposal No.: 2022-05321
• Project Start Date: Sep 15, 2022
• Project End Date: Sep 14, 2026
• Grant Year: 2022
• Program Code: [SCRI]- Specialty Crop Research Initiative

Recipient Organization
OHIO STATE UNIVERSITY
1680 MADISON AVENUE
WOOSTER,OH 44691

Performing Department
(N/A)

Non Technical Summary
Controlled environment agriculture (CEA) means crop production under growing conditions created in greenhouse or other types of systems so that greater crop yield can be obtained. Relative to conventional field production, CEA also improves crop health, enhances product quality, extends the growing season, and reduces resource input and waste, while maintaining high productivity. CEA for specialty crops increased rapidly over the past 20 years in the US following a worldwide trend. Successful environment control and crop management in CEA requires equipment, data, and grower experience. However, a shortage of experienced growers (who make key production and management decisions) often slows down development across the CEA industry. CEA operations, especially in greenhouses, require broad knowledge of and experience in crop production and engineering systems management. Because of the limited pool of experienced growers, many in the greenhouse industry still rely on the human resources "imported" from outside the US (e.g., Mexico, Canada, and European countries). Therefore, the US needs to develop its own talent pool of experienced growers and/or technologies that can empower growers with limited experience.In addition to grower shortages, CEA technology development by domestic R&D capacity in the US largely lags behind. Historically, new technologies have been imported from the Netherlands, the world leader in CEA. This heavy dependence on overseas technologies in a rapidly expanding industry creates unbalanced accessibility and affordability depending on the scale of operations. Differences in industry scale, diversity, design standards, measurement unit systems, and government regulations also contribute to delays in the adaptation of new technologies in the US from overseas. US-based innovations and technology development are critical for the sustainable development of the CEA industry in our country.CEA growers use various sensors and measurements to understand the environmental conditions as well as crop physiological status to maintain optimum growing conditions. Recent technology developments expanded the range of sensors available for greenhouse growers; but their effective use is still limited. Often growers make decision without knowing exactly the environmental spatial/temporal uniformity, crop physiological status, and real-time resource consumption. One reason could be the lack of an integrating platform for sensors (wired or wireless) and user data (e.g., additional sensors and plant growth measurements conducted manually by growers). Typically, various systems are installed (because of lack of integration) resulting in redundancies and cumbersome software configurations, increasing production costs. An integration platform is needed that is compatible with new algorithms to analyze data so that growers can better optimize productivity and profitability. For example, the application of artificial intelligence (AI) has been proposed as a next technical milestone for the CEA industry. The use of AI can assist growers to make more logical, site-specific data-based management decisions. These management decisions largely influence crop productivity, quality, as well as usage of labor and other resources. An AI combined with greenhouse expert knowledge can be used to optimize decisions for the control of the greenhouse. We plan to use the so-called digital twin technology (a.k.a. virtual models) to better understand and simulate the complex plant growing environment. The unique collaborative effort engages both industry and academia as the project team (Ohio State University, Rutgers University, Koidra, Cornell University, and University of Arizona). We aim to develop core components for more advanced climate control and crop management practices by integrating sensing, modeling, and optimization of greenhouse crop production across the varied range of growing systems and across various climate zones.The proposed project has social significances as our outcomes will contribute to: 1) Reducing our country's dependence on the imported innovations and human resources in CEA by creating an innovative technological platform, data-driven decision making strategies, and workforce training opportunities; 2) Improving energy and water efficiency of CEA operations resulting in a lower environmental impact. Our long-term goal is to make the US CEA industry more profitable and sustainable as a key component of US food production. The project will have four specific key objectives to achieve the goal.Objective 1) Development of a data- and model-driven decision-making platform We will use wireless sensors to collect environmental and plant data. Once the data are transmitted, they can be accessed by the project members using an internet connection. We will expand data processing and developing CEA system simulation models. The sensor data will be delivered to a cloud-based management system to be processed further. A machine-learning (ML) framework will be developed using physics-informed deep learning. We will develop a decision support system for CEA operations, leveraging the sensor systems and the AI-based digital twin.Objective 2) Validating the efficacy of new data- and model-driven decision makingA side-by-side comparison with conventional management practices will be performed using modern, "high-tech", Dutch Venlo style greenhouses (OH), "medium-tech" glass greenhouses (NJ), "medium-tech" inflated double-layered polyethylene film covered greenhouses (AZ), and "low-tech" unheated simple structures (OH) over multiple seasons for tomato and lettuce crops. Resource use (electric power, heating fuel, cooling, irrigation water, and labor) will be compared in addition to crop yield and quality.Objective 3) Understanding socioeconomics of greenhouse technology adaptationWe will use a three-phased approach based on available knowledge about CEA industry sector and its technological market dynamics. We will evaluate the current decision-making process of greenhouse growers. We will estimate economic feasibility of new technology adoption working with our stakeholder advisory group. Finally, we will present effective business strategies for introducing new technologies.Objective 4) Engaging our stakeholders through professional learning opportunities for workforce developmentWe will develop training modules for growers and greenhouse managers that will be available for use by universities and businesses. In addition, we will take a multi-step approach to engage with various stakeholders and organize regular events to discuss strategies and future directions of our R&D efforts. Workshops, webinars, and discussion forums will engage growers. Learning opportunities of core engineering principles (sensing, modeling, and control) will be offered to our grower stakeholders.Our systems approach to advance US based CEA technologies will contribute to the USDA SCRI focus areas of 'New Innovation and Technology Development' and 'Increasing Crop Productivity and Production Efficiency'. Obj. 1 will result in a novel AI-based greenhouse climate controller with input from academic and industry expertise in computer science, crop production, and greenhouse engineering. Obj. 2 will validate the novel control approach in different production systems and climate conditions. Obj. 3 will evaluate the socioeconomic considerations involving technology adaptation by stakeholders (growers and supply companies), and Obj. 4 will engage stakeholders by providing learning opportunities that contribute to workforce development. Altogether, our proposed project will bring the US CEA industry to the next level, allowing it to develop as a global leader.

Animal Health Component

50%

Research Effort Categories

Basic

20%

Applied

50%

Developmental

30%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
205	2410	1020	20%
404	2410	2020	30%
601	6299	3010	20%
404	2410	2080	30%

Knowledge Area
601 - Economics of Agricultural Production and Farm Management; 205 - Plant Management Systems; 404 - Instrumentation and Control Systems;

Subject Of Investigation
2410 - Cross-commodity research--multiple crops; 6299 - Marketing, general/other;

Field Of Science
2020 - Engineering; 2080 - Mathematics and computer sciences; 3010 - Economics; 1020 - Physiology;

Keywords

artificial intelligence

greenhouse

modeling

optimization

sensing

Goals / Objectives
Successful controlled environment agriculture (CEA) operations need technologies and growers who can operate the technologies effectively with broad knowledge of and experience in climate and crop management in controlled environments. While the acreage of CEA has been expanding in the US, most technologies and growers employed in the US CEA are imported from other countries. CEA technology development through domestic R&D capacity in the US lags. Unique characteristics of the US greenhouse industry also slow down the adaptation of most advanced technologies. In addition, differences in design standards, measurement units and government regulations contribute to delays in the adaptation of new technologies in the US. Therefore, US-based innovations and technology development are critical for the sustainable development of the US CEA industry. Our long-term goal is to make the US CEA industry more profitable and sustainable, and less reliant on foreign technologies. Technological advancements to better manage CEA crop production is critical. We believe that the integration of artificial intelligence (AI) is a critical next milestone for the CEA industry. Worldwide, advanced CEA operations have a well-established infrastructure of computer-based greenhouse climate control systems. Integration of AI can assist growers in making more logical, data-driven, site-specific management decisions assisted by a foundational knowledge base including engineering, crop physiology, and economics. These management decisions largely influence crop productivity, quality, as well as use of labor and other resources. An AI framework consists of models, controllers, and real-time data, that combined with domain knowledge result in optimized decisions for selected outcomes such as profitability and resource use efficiency. Advancing this concept requires more R&D efforts involving alliances of computer scientists and application specialists. Recent technology developments expanded the range of sensors available for growers in various scales and systems of CEA. However, growers often make decisions without knowing exactly the environmental spatial/temporal uniformity, crop physiological status, and real-time resource consumption. Our project team engages both industry and academia and aims to develop core components ready to use for more advanced climate control and crop management practices. Our proposed project has the following four specific objectives.Objective 1: Development of a data- and model-driven decision-making platformObjective 2: Validating the efficacy of new data- and model-driven decision makingObjective 3: Understanding the socioeconomic aspects of greenhouse technology adaptationObjective 4: Engaging our stakeholders through professional learning opportunities that contribute to workforce development.

Project Methods
Objective 1) Development of a data- and model-driven decision-making platform Task 1A: Using WSNs to collect environmental and plant dataSensor cluster readings as well as digital images of plant status will be wirelessly transmitted to a gateway connected to a computer that has access to the internet. The sensor clusters will be installed at multiple locations throughout the growing area so that the spatial and temporal variation of the readings can be assessed.Task 1B: Expanding data processing and developing CEA system simulation modelsData processing: Data from sensors will be analyzed and visualized through a cloud-based GUI for easy access by growers. The data will also be used to predict future plant growth and environmental conditions. We will use a model-based reinforcement learning (MBRL) approach for the remote control of the greenhouse environment. Within the MBRL framework, AI models consist of three modular components: Data aggregator, Digital twin (model), and Controller.Model training, testing and validation: The simulation and decision models will be trained and validated using available and published data sets. We will continuously refine and retrain the models as well as improve the code during this validation process. The accuracy of the overall approach will be cross-validated against real-time sensor data. Our modeling work will be evaluated using several different techniques: validation based on historical data, parameter sensitivity analysis, and predictive evaluation. Collected plant growth and environmental data will be evaluated using a variety of statistical methods including analysis of variance, regression analysis, and time series analysis.Task 1C: Decision making support through optimizationWe will develop an AI-based autonomous optimization strategy using nonlinear MPC and economic MPC that can satisfy production targets. Output from the optimization strategy will: 1) be returned to the digital twin model (Task 1B) for additional simulation and verification, and 2) be sent to the greenhouse control system for implementation.Objective 2) Validating the efficacy of new data- and model-driven decision makingSide-by-side comparisons with conventional management practices will be performed using modern, "high-tech", Venlo-style greenhouses (OH), "medium-tech" glass greenhouses (NJ), and "medium-tech" inflated double-layered polyethylene film covered greenhouses (AZ) over multiple seasons for tomato and lettuce crops. Resource use will be compared in addition to crop yield and quality. At each research site, the local project team will examine the new approach focusing on improving specific aspects of crop performance and resource use efficiency. We will also examine a ML approach to forecast the near-future temperature in high tunnels using "low-tech" unheated simple structures (OH).Objective 3) Understanding socioeconomics of greenhouse technology adaptationWe will use a three-phased approach based on available knowledge about the CEA industry sector and its technological market dynamics. Phase I will map the decision-making process used by greenhouse growers and investors to contract builders, technology providers, and consultants to assist in construction projects. Phase II will estimate economic feasibility of AI technologies in commercial greenhouses. Phase III will propose marketing strategies that US-based technology providers may use to structure persuasive service offers to different groups of CEA growers and investors.Objective 4) Engaging our stakeholders through professional learning opportunities that contribute to workforce developmentWe will develop nationwide CEA engineering extension capacity and training modules for growers and other CEA professionals. The training modules planned currently are a total of 22, each consisting of an annotated slide deck covering a 20-30 min talk, review questions, and additional resources. They include basic and advanced topics of greenhouse engineering, artificial intelligence, data analysis, crop physiology, and environmental control systems. Additional supplemental materials such as assessment questions and instruction guides will be developed. Selected modules will be packaged as professional courses over the project years and beyond. Participants who complete the entire educational program will obtain a 'certification of completion.' In addition, we will take a multi-step approach to engage with various stakeholders and organize regular events to discuss strategies and future directions of our R&D efforts. Workshops, webinars, and discussion forums will engage growers/professionals. Learning opportunities of core engineering principles (sensing, modeling, and control) will be offered.Project management and evaluations. Our project's organizational structure consists of the Project Leadership and Management Team, Project PI Teams (Objective teams), Focused Working Groups and the Advisory Board. The Leadership and Management Team will frequently interact with all project members and the members of the Advisory Board. The Leadership/Management Team's responsibilities include providing guidance and oversight with respect to project deliverables, making decisions when progress falls behind the project's timeline, being a cheerleader for the entire project team, and making sure that feedback from the Advisory Board is carefully considered and implemented. The Project Manager will be responsible for handling day-to-day project administrative requirements, coordinating meetings, scheduling outreach events, and assisting the PDs/PIs with the writing of annual progress reports.In addition to the Advisory Board, we plan to form Focused Working Groups for specific areas (e.g., sensing/instrumentation, smart crop production, economics, and workforce development) which will contribute to and critically review the outcomes of our project. Each Focused Working Group will have at least 2-3 stakeholder members. A project PI will be assigned as the coordinator for each group by the PD Kubota and hold at least biannual meetings to discuss strategies and progress and, in some cases, to redirect our R&D efforts as appropriate. Annual project progress reports will be evaluated by the Advisory Board members who will provide written feedback after each meeting. In some cases, non-disclosure agreements will be initiated with Focused Working Group and Advisory Board members and trial participants so we can share proprietary and/or unpublished scientific information.

Progress 09/15/23 to 09/14/24

Outputs
Target Audience:Our target audiences include CEA growers and operators, applied computer scientists, venture capital investors, equipment manufacturers of sensors, climate controllers and other CEA equipment, greenhouse consultants, agricultural engineers, university faculty and Extension personnel, and undergraduate and graduate students. In particular, our team made the effort to reach out to audiences who need to develop foundational knowledge about greenhouse climate control through webinars, workshops, symposia, and online courses. Changes/Problems:The challenges of clearly defining and maximizing profits for an objective function became evident at CU where optimizations were attempted. While we have made strides in demonstrating energy savings and improved efficiencies, ensuring maximum profitability remains a challenge that we are actively working to address. We have experienced delays in integrating Koidra's AI into the existing greenhouse climate control systems at UA and RU research greenhouses. Delays occurred due to the challenges of integrating new equipment with already existing (and older) equipment, sending data and control settings to a third party (Koidra) and receiving new control settings from that third party, supply chain delays, and compatibility issues. A recent software update and a limited number of available IP addresses in one of the research facilities caused communication challenges between the local control system (Argus) and the remotely operating AI strategy (Koidra). In addition, technical services of climate controllers were not always consistent, suggesting that better communication and collaboration among technology providers are needed. We will continue our efforts to resolve these challenges. Designs and control systems specific to university greenhouses also caused unexpected issues and delays. OSU has a modern greenhouse complex that employs technologies and equipment similar to what can be found in modern commercial greenhouses. However, due to the smaller footprint of the OSU research greenhouse, some technologies not widely used in large-scale commercial greenhouses are used for achieving target climate conditions as specified by faculty researchers. One example is the use of a mechanical ventilation system with evaporative cooling (the so-called pad-and-fan cooling system). Koidra had to spend a significant amount of time to better understand the system's behavior and then integrate this cooling method into their digital twin and the climate control algorithm. While design and operational standardization is common in greenhouses in the Netherlands, U.S. greenhouse designs and control strategies can differ substantially, making it more challenging to come up with a one-size-fits-all approach. The challenges associated with this diversity in greenhouse design and operation will have to be further addressed by the project team and will be discussed with the newly proposed industry community. An AI-based climate control system that partially operates on the internet requires steady network communication between the climate control server and the greenhouse control panels. During our research, we have encountered network communication errors associated with brief disruptions of system communications. This could become a more universal challenge when dynamic control and optimization are attempted using the existing climate controller installed at commercial CEA operations. Infrastructure improvements are not always easy or cheap and therefore the project team will discuss ways to minimize the impact of temporary network communication errors. The integration of additional sensors (e.g., canopy temperature sensor, RGB camera, CO2 sensor) within the existing climate control system has taken a longer time than anticipated at RU. The research greenhouse staff has limited bandwidth to assist with sensor and communication equipment installation. We are currently working on resolving these issues or adjust our plans accordingly. What opportunities for training and professional development has the project provided?The ADVANCEA Project offers a unique opportunity for researchers and technicians to interact with decision-makers across the CEA industry and to investigate novel strategies pertinent to greenhouse climate control. Involvement in the project offers unparalleled opportunities for learning and professional development. The ADVANCEA project actively engages our graduate students and postdocs during project related meetings. Monthly project meetings and objective-specific discussion meetings are attended by 3-5 graduate students/postdocs regularly. We also provided extracurricular training opportunities through our ADVANCEA Academy course to a total of 17 graduate students/postdocs to learn about greenhouse climate control and crop management. In addition, a number of undergraduate students are involved as research assistants at the University of Arizona (3), Rutgers University (2) and Ohio State University (6) to conduct greenhouse experiments, through which they learn about greenhouse systems, hydroponics, fertigation system, sensors and controls. This inclusion of undergraduate students in research programs to develop new technologies is particularly important for attracting new students to the CEA industry. The project also provided project member's students and employees with significant professional growth, expanding their knowledge in AI and machine learning by developing an autonomous, deep-learning-based control system. They have honed their skills in data analysis and model development. Attending meetings and conferences has been pivotal in keeping the team abreast of the latest technological advancements. For example, wireless sensor installation experiences improved handling IoT applications within greenhouse settings. Collaborations between industry and academia have fostered a valuable exchange of knowledge, enriching both parties in the fields of advanced horticultural technology and science. The work ongoing at RU is providing hands-on training for undergraduate and graduate students. The work at RU is also informing the project team members about the challenges of integrating a remotely-operated AI control system into an existing on-site stand-alone computer control system that was purchased from one of the better-known greenhouse control vendors (Argus). How have the results been disseminated to communities of interest?During the reporting year, we wrote four journal articles, five Extension/growers articles and gave 14 presentations. Several more manuscripts are under preparation and will be published in the following year. We expanded a project website with more educational contents relevant to the CEA industry. We reached many professionals currently active in controlled environment agriculture through workshops, conferences, events, and short courses, as described under the Objective 4 accomplishments. Tours to our greenhouse research facilities facilitated outreach to a wide range of audiences including the public-at-large, students, and industry professional, showcasing technologies for advanced crop production in CEA facilities. Compared to greenhouse horticulture and plant science, greenhouse engineering education in the U.S. has been relatively weak. Enhancing our project's visibility through publications and tours is a significant effort for members of the ADVANCEA project team. What do you plan to do during the next reporting period to accomplish the goals?Objective 1: Following the successful integration at OSU, Koidra will deploy both Opera and AI-control systems at UA and RU. Koidra will assist with AI supported growing trials, involving tomato (UA) and lettuce (RU), respectively.Working with OSU, Koidra plans to fine-tune its algorithm for irrigation control using a digital twin at OSU. Koidra will develop their model predictive controller (MPC) framework for all three collaborating institutions.As a parallel effort, UA plans to further advance its irrigation control based on cumulative solar radiation and drainage rate, evaluating water and energy savings compared to time-based controls. CU plans to work on: 1) energy optimization for networked greenhouses using physics-informed neural networks and predictive control, 2) decarbonization of greenhouse systems by using waste energy from bitcoin mining and model predictive control, and 3) evidential learning-based model predictive control for greenhouse climate. The transition to networked greenhouses has opened new possibilities for leveraging data-driven technologies to optimize resource allocation and environmental conditions. Physics-informed neural networks, integrated with model predictive control, present a promising solution. This approach not only optimizes energy use across connected greenhouses but also ensures that operational decisions are informed by accurate simulations of physical processes, leading to more sustainable agricultural practices. UA will fine-tune MPC strategy for greenhouse climate control and continues collecting 2nd year data from wired and wireless sensor network in the project greenhouse on crop and greenhouse climate variables.RU will work with other ADVANCEA team members to design lettuce experiments that can be used to inform crop model development efforts that will improve the environmental control decisions made by the AI -based control system developed by Koidra. Objective 2: Following the successful development of pruning strategies, OSU will initiate developing data-driven strategies for nutrient management. We will evaluate a new approach of balancing plant growth and nutrient dosing, compared with the conventional strength- based (ppm or EC) management practices. We will also integrate our wireless light sensor installed at the bottom of the tomato canopy to quantify the leafiness as a feedback for irrigation control. RU will conduct a series of hydroponic lettuce experiments designed to provide information about crop performance (growth rate), crop quality (minimizing tipburn), water consumption, and energy consumption (for greenhouse heating, cooling, and lighting). UA continues collaborating with Koidra to evaluate the AutoPilot platform and AI integrated controls.They will also collaborate with ADVANCEA members on crop and greenhouse climate model development and validating efforts. Objective 3: OSU continues a qualitative content analysis of interview results that will develop a comprehensive description of the investment decision journey, including the identification of discovery channels, sales approaches by technology and service firms that are more likely to culminate in transactions (as opposed to non-sale closings), marketing touchpoints with positive evaluations, and offer elements (factors) that enhance the probability of parties to establish long-term commercial relationships. In the second stage of the project, OSU will pivot to the financial aspects of greenhouse technology adoption. Re-connecting with study participants, OSU will construct enterprise budgets for representative operations and create automated spreadsheets for simulation and sensitivity analysis. This effort will include cash flow and feasibility analyses for alternative technologies being examined in other project objectives. The final stage of objective 3 will evaluate the perceptions of CEA production companies' executives regarding alternative offering schemes. The results found in stages one and two will examine the executives' propensity to adopt a given technology when it is framed under alternative structures. Objective 4: UA, RU, and OSU will organize the second offering of our ADVANCEA Academy online course 'Introduction to Greenhouse Environmental Control for Crop Production' with revised/improved contents during January - April 2025. As a parallel effort, we will begin to design and plan to offer additional courses that discuss critical contents for CEA professionals to master greenhouse engineering and technology. We will continue to deliver monthly webinars as part of Indoor Ag Science Café, collaborating with other USDA-NIFA SCRI funded project teams (OptimIA and CEA HERB). As new initiative, we will plan to develop a communication network for U.S. industries providing technologies relevant to CEA, especially focused on sensing systems and climate controllers. Each PI will continue being active in their own Extension programs, including UA's 24th Annual Greenhouse Crop Production and Engineering Design Short Course, and OSU's Leafy Green Crop Event, and OSU's Annual OHCEAC Conference. These events will include topics contributing to professional development for members of the U.S. CEA industry.

Impacts
What was accomplished under these goals? Objective 1: The Ohio State University (OSU) started using Koidra's system. The wireless sensor system including a thermal camera were connected using a LoRaWAN gateway to the OSU network. The collected sensor data were incorporated into Koidra's decision making algorithm to change greenhouse climate control set points. Koidra completed the development of a digital twin model framework for the OSU greenhouses. The model was trained using historical data at OSU. OSU examined Koidra's new interface Opera, which enables users the flexibility to develop control logic through an intuitive, low-code approach. Koidra examined the newly developed control logic and incorporated it in its AI based physics-informed deep learning framework. The University of Arizona (UA) collected data from their wired and wireless sensor network in the project greenhouse on crop and greenhouse climate variables. Wired sensors deployed in each of the three zones in greenhouse were for irradiance, PAR, air temperature, relative humidity, and crop temperature. As part of the crop fertigation system, electrical conductivity and pH of the nutrient solution were also collected. Drainage percentage and daily crop water consumption were monitored. Wireless sensors were for PAR, air temperature and relative humidity, along with crop canopy temperature. A database has been created for data collection and storage, and it is available for model validation studies. At Cornell University (CU), simulation models of CEA systems were developed, including integrated rooftop greenhouses (i-RTGs), networked greenhouses, and plant factories with electric lighting. These models allow for the prediction and optimization of the crop environment. The potential benefits of i-RTGs on top of buildings to reduce energy and CO2 costs were investigated. Simulations with 11 urban locations across the U.S. found that integrating i-RTGs with buildings under the nonlinear model predictive control framework resulted in a 15.2% decrease in energy and CO2 costs. Furthermore, the i-RTG model proved adaptable to various climates, with colder regions showing more significant cost-saving potential. This research highlights the benefits and feasibility of incorporating i-RTGs on top of buildings, offering a sustainable, decarbonizing strategy for urban agriculture and building management. Objective 2: OSU continuously examined a new tomato leaf pruning strategy driven by the PAR intensity at the lower leaves of plant canopy. Leaf pruning is a labor-intensive practice typically conducted weekly in commercial greenhouses. Preliminary data from the first experiment in 2023 showed that the frequency of pruning and degree of leaf removal could be optimized based on the weekly integration of PAR measured near the bottom of the plant canopy. An experiment conducted from January through April 2024 confirmed the efficacy of our new data-driven pruning methodology in reducing labor used for leaf pruning (40% reduction over 14 weeks). UA designed and implemented a three-zoned crop production system with autonomous drip irrigation to grow a tomato crop in the project greenhouse. An autonomous irrigation strategy was created based on accumulated solar radiation and drainage rate-based control and compared it to traditional time-based irrigation control. It was demonstrated that the accumulated solar radiation based control was a more precise and dynamic plant needs-based control strategy compared to time-based control. UA developed and evaluated a nonlinear autoregressive artificial neural network (NARX) integrated with an MPC (model predictive control) strategy for greenhouse climate control. The control was based on estimating the greenhouse air temperature and optimizing control actions for the exhaust fans, evaporative cooling pad pump, and energy curtains. The model and control routine were tested, comparing the control actions and energy consumption between the MPC and an on-off control system. On average, the conventional on-off control showed a 2°C difference from the setpoint, while the MPC reduced this to 1.4°C. Energy consumption was higher with the MPC when the outside temperature was close to the setpoint, but lower when the external temperature was above the setpoint, resulting in an estimated 15% energy savings in these cases. Rutgers University (RU) installed a hydroponic NFT (nutrient film technique) systems in the NJAES Research Greenhouse, and conducted the first hydroponic lettuce trial. The greenhouse system capabilities include heating (hot-water system), cooling (mechanical ventilation with evaporative cooling pads), supplemental lighting (high-pressure sodium fixtures), shading, and vertical airflow fans (one over each NFT system). The experimental setup includes two identical NFT systems, each with 14 troughs holding 18 lettuce plants per trough, and each system with its own nutrient solution delivery system (holding tank, recirculation pump, and plumbing). Objective 3: OSU conducted interviews with 28 selected CEA organizations to examine the investment decision journey for new technology purchases. By describing discovering channels, marketing touchpoints, and critical factors leading to decisions, this objective will assist U.S. technology and service providers with framing sales approaches with a high probability of purchases and with establishing long-term relationships with CEA producers and suppliers. Objective 4: Through the ADVANCEA Academy (https://www.ceaforum.org/advancea-academy) a 13-week online course (26 lectures) 'Introduction to Greenhouse Environmental Control for Crop Production' was offered during January-March, 2024. A total of 138 people registered to master core concepts of greenhouse engineering. Additionally, we organized a half day workshop 'Greenhouse Climate Control - Sensors and Control Strategies' as part of tHRIve (Horticulture Research Institute) symposium on July 13, 2024. Approximately 100 people attended the workshop. An open house event was held at OSU on July 13, 2024 to showcase the ADVANCEA project. Approximately 40 participants attended this event and learned about advanced climate control for crop production in greenhouse. UA and OSU organized two grower short courses: 1) 23rd UA-CEAC Annual Greenhouse Crop Production and Engineering Design Short Course, and 2) 2024 Greenhouse Management Workshop. As part of the UA-CEAC short course, an Artificial Intelligence panel led by Murat Kacira (UA) was organized with expert panelists. A Model Predictive Greenhouse Control approach was demonstrated to more than 80 participants during the short course. OSU organized a symposium '3rd OHCEAC Annual Conference' that focused on advancements in automation and crop management. Participants were 103 in person and 84 online to learn greenhouse automation, AI-based climate control, horticultural lighting, and data-driven crop and pest management. ADVANCEA outreach events also targeted public audiences and K-12 school children in addition to professionals. Our project sites were visited by various local, national, and international visitors, students, and groups of professionals, showcasing ongoing research projects is a highlight. During the reporting period, the OSU facilities were introduced to a total of 8,818 visitors. UA also provided tours to a total of 300+ visitors to their greenhouse research facilities. We formed a partnership with other USDA-NIFA SCRI funded project teams (OptimIA and CEA HERB) to conduct a monthly webinar series (Indoor Ag Science Café) beginning in August 2024. This webinar series was originally developed by OptimIA and has more than 1,300 subscribers, but the focus was on indoor vertical farming. With the participation of ADVANCEA and CEA HERB, the scope was expanded to include all CEA systems, which will expand the target audience.

Publications

• Type: Peer Reviewed Journal Articles Status: Published Year Published: 2024 Citation: Ajagekar, A., B. Decardi-Nelson, and F. You. 2024. Energy management for demand response in networked greenhouses with multi-agent deep reinforcement learning. Applied Energy 355:122349.
• Type: Peer Reviewed Journal Articles Status: Published Year Published: 2024 Citation: Chen, W.H. and F. You. 2024. Decarbonization through smart energy management: Climate control in building-integrated rooftop greenhouses for urban agriculture. Journal of Cleaner Production 458:142544
• Type: Peer Reviewed Journal Articles Status: Published Year Published: 2024 Citation: Decardi-Nelson, B. and F. You. 2024. Artificial intelligence can regulate light and climate systems to reduce energy use in plant factories and support sustainable food production. Nature Food 5:869-881.
• Type: Peer Reviewed Journal Articles Status: Published Year Published: 2024 Citation: Hu, G. and F. You. 2024. AI-enabled cyber-physical-biological systems for smart energy management and sustainable food production in a plant factory. Applied Energy 356:122334.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Both, A.J. 2024. Measuring and controlling light. tHRIve Symposium Greenhouse Climate Control: Sensors & Control Strategies. Cultivate 2024, July 13, 2024.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Chen, W.H. and F. You. 2023. Model predictive control on integrated-rooftop greenhouse climate control. 26th conference on Process Integration, Modelling and Optimisation for Energy Saving and Pollution Reduction - PRES'23, 2023, vol. 103, pp. 109-114.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Chen, W.H. and F. You. 2024. MPC for the indoor climate control and energy optimization of a building-integrated rooftop greenhouse systems. 12th IFAC Symposium on Control of Power and Energy Systems - CPES 2024, 2024, vol. 58, no. 13, pp. 164-169.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Decardi-Nelson, B. and F. You. 2023. Improving resource use efficiency in plant factories using deep reinforcement learning for sustainable food production. 26th conference on Process Integration, Modelling and Optimization for Energy Saving and Pollution Reduction - PRES'23, 2023, vol. 103, pp. 79-84.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Decardi-Nelson, B., A. Ajagekar, and F. You. 2024. Multi-agent deep reinforcement learning for energy management in grid-responsive networked greenhouses. 2024 American Control Conference (ACC), pp. 129-134.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Hu , G. and F. You. 2023. Energy management for plant factory with deep learning and predictive control. 26th conference on Process Integration, Modelling and Optimisation for Energy Saving and Pollution Reduction - PRES'23, 2023, vol. 103, pp. 91-96.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Hu, G. and F. You. 2024. Assessment of AI-based robust model predictive control application in large-scale photovoltaic-based controlled environment agriculture for urban agriculture. 12th IFAC Symposium on Control of Power and Energy Systems - CPES 2024, 2024, vol. 58, no. 13, pp. 368-373.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Kacira, M. 2024. Measuring and controlling temperature. tHRIve Symposium Greenhouse Climate Control: Sensors & Control Strategies. Cultivate 2024, July 13, 2024.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Kacira, M. 2024. Resource Use Efficient Controlled Environment Agriculture Systems. XV Forum on Research and Advances. Autonomous University of Chapingo, March 18th
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Kim, C. 2024. Development of a data-driven leaf pruning method based on weekly light integral below the canopy. 2024 OHCEAC Annual Conference, July 17, 2024, Columbus, OH.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Kim, C. and C. Kubota. 2024. Theoretical evaluations of a data-driven tomato leaf pruning method based on weekly light integral at lowest leaf. Presentation at International Symposium on Light in Horticulture, May 19-22, 2024, Seoul, Korea.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Kubota, C. 2024. Measuring and controlling EC, pH, DO, and other rootzone parameters. tHRIve Symposium Greenhouse Climate Control: Sensors & Control Strategies. Cultivate 2024, July 13, 2024.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Kubota, C., A.J. Both, and K. Satoh. 2024. A tradeshow for CEA innovations. Presentation at Indoor Ag Science Caf�, August 27th, 2024. https://www.scri-optimia.org/showcafe.php?ID=111208
• Type: Websites Status: Published Year Published: 2024 Citation: Kubota, C.. 2023. VPDleaf vs. VPDair  Two different ways to determine VPD. eGro Edible Alert. 8:16. https://www.e-gro.org/pdf/e816.pdf
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Ling, P. 2024. Measuring and controlling humidity and CO2. tHRIve Symposium Greenhouse Climate Control: Sensors & Control Strategies. Cultivate 2024, July 13, 2024.
• Type: Websites Status: Published Year Published: 2024 Citation: Satoh, K., C. Kubota, and A.J. Both. 2024. Visiting GPEC tradeshow in Japan. https://www.ceaforum.org/visiting-gpec-tradeshow-in-japan
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Tran, K. 2024. Using artificial intelligence for greenhouse climate control. tHRIve Symposium Greenhouse Climate Control: Sensors & Control Strategies. Cultivate 2024, July 13, 2024.
• Type: Websites Status: Published Year Published: 2024 Citation: Tran, K., A.J. Both, and C. Kubota. 2024. A primer of artificial intelligence for greenhouse control. eGro Edible Alert 9:8. https://www.e-gro.org/pdf/e908.pdf

Progress 09/15/22 to 09/14/23

Outputs
Target Audience:Our target audiences include greenhouse and indoor farming growers and operators, applied computer scientists, venture capital investors, R&D personnel for sensors, climate controllers and other CEA equipment, greenhouse consultants, agricultural engineers, university faculty and extension personnel, and undergraduate and graduate students. Especially the team made the effort to reach out to audiences who need to develop foundational knowledge on greenhouse climate control and developed plans of an in-person workshop and an online course to be offered in 2024. Changes/Problems:Our project officially started on September 15th, 2022; however, PIs experienced some delays in hiring project personnel (Ohio, Rutgers) and in establishing subcontracts. However, these are minor delays, and we should be able to accomplish project outcomes as planned within our proposed timeline. Koidra and Cornell originally proposed to closely work together to develop an optimization and decision making platform. However, due to IP issues, we re-arranged the dynamics and maintain the communication between these two groups of computer scientists and modelers to identify ways to integrate different optimization frameworks. PDs will assist with communication, monitor progress and provide guidance as needed. We experienced some delays in securing greenhouse spaces and institutional IT support. Introducing AI and remote processing into existing institutional greenhouse control systems that serve multiple users at individual institutions needed longer discussion and communication to achieve better understanding by the greenhouse managers and institutional IT support teams. However, through increased communication, we are achieving better understanding and we hope to make more rapid progress during the 2nd year. OSU's research started in a brand new greenhouse facility and some equipment needed repairs due to design/installation issues. Consequently, it took longer for our greenhouse manager, research personnel, and PI to be able to use the new greenhouse facilities as intended. The greenhouse facility's staffing experienced some delay. The horticultural and engineering staff members started work assisting the greenhouse manager in November 2023, which will be a big help for project management during the 2nd year and beyond. What opportunities for training and professional development has the project provided?The ADVANCEA project actively engages our graduate students and postdocs during project related meetings. Monthly project meetings and objective-specific discussion meetings are attended by 3-5 graduate students/postdocs regularly. In addition, a number of undergraduate students are involved as research assistants at the University of Arizona (3), Rutgers University (1) and Ohio State University (6) to conduct greenhouse experiments through which they learn about greenhouse systems, hydroponics, fertigation system, sensors and controls. This inclusion of undergraduate students in research programs to develop new technologies is particularly important for attracting new students to the CEA industry. The ADVANCEA project offers a unique opportunity for our project economist (Signorini and his staff technician) to interact with decision-makers in the CEA industry and investigate using qualitative methods pertinent to the research objectives. The daily project activities have served as an unparalleled opportunity for professional development. At Koidra, the project has afforded the team valuable professional development through the creation of the Opera framework, enhancing their capabilities in developing low-code decision-making platforms. They have also expanded their expertise in AI and machine learning while developing an autonomous, deep-learning based control framework. The necessity of installing wireless sensors has enhanced experience in IoT implementation in greenhouse environments, while partnerships with academic institutions have facilitated a rich, reciprocal knowledge exchange in advanced horticultural technology and science. At Cornell, the ADVANCEA project has provided team members with the opportunity to immerse themselves in AI techniques related to CEA. They have honed their skills in data analysis and model development. Attending international conferences has been pivotal in keeping the team abreast of the latest technological advancements around the world. How have the results been disseminated to communities of interest?During the first project year, we wrote two journal articles and gave 12 presentations. We launched a project website and developed educational content relevant to the CEA industry. Tours to the greenhouse research facility facilitated outreach to a wide range of audiences including the public-at-large, students, and industry professionals. During our first year, we reached a large number of professionals currently active in controlled environment agriculture through workshops, conferences, events, and short courses. We have also offered tours to a wide range of visitors showcasing technologies for advanced crop production in CEA facilities. Specifically, OSU brought a large number of visitors to the CEA research complex where our ADVANCEA project was highlighted during tours that included 284 CEA professionals, 474 stakeholders, and 3,780 public/students. Arizona gave tours to approximately 150 professionals to introduce the project, the environmental monitoring and control system, and hydroponic greenhouse production. The total number of visitors to these facilities are estimated to be 4,690. What do you plan to do during the next reporting period to accomplish the goals?Due to the delays in starting the project, the majority of main activities under the four project objectives will start during Year 2. With a new project coordinator hired in September 2023, efficiency of our project management will be improved to meet the expectations of all members. Specific plans for future accomplishments by the individual PI groups are described below. Objective 1: Development of a data- and model-driven decision-making platform Koidra will deploy both Opera and AI-control systems at OSU, UAz and Rutgers. In particular: Between Nov 2023 and April 2024, Koidra aims to deploy Opera at OSU and UAz for irrigation control and other hard-to-automate climate control tasks - while allowing OSU and UAz to concurrently carry out the leaf pruning strategy experiment. Starting July 2024, Koidra will deploy AI framework in one OSU greenhouse to optimize the greenhouse climate to maximize the crop yield while minimizing resource consumption, and comparing results with another greenhouse managed with a conventional control approach. At UAz and Rutgers, Koidra also aim to run AI supported growing trials, involving tomato and lettuce, respectively. Cornell will delve deeper into both the efficiency and the profitability aspects of our research. While energy savings and sustainability are paramount, the success of any agricultural endeavor ultimately hinges on its profitability. The AI models will be further refined based on feedback and outcomes from real-world applications. The use of the models will be expanded to encompass more diverse climates and crops. Building on our existing AI models, we will develop optimization models specifically aimed at maximizing profitability. These models will factor in variables like energy costs, market prices for crops, and other operational expenses to recommend the most cost-effective operational strategies. Objective 2: Validating the efficacy of new data- and model-driven decision making OSU will further evaluate the new data-driven leaf pruning strategy by integrating it into Koidra's AI based climate and crop management strategies. Rutgers will finish installation of the experimental setup (greenhouse facility layout, growing systems, sensors, data acquisition) to conduct lettuce growth trials using a nutrient film technique system. Arizona will continue growing a tomato crop, collect crop and greenhouse environment related data, along with water and energy use data. The collected data will be analyzed for model development and validation studies. As for low-tech greenhouses, OSU will focus on evaluating the methods of combining a deterministic model with the deep learning (DL) model for high tunnel climate predictions. The process will involve developing a more advanced neural network architecture and exploring recursive learning functions to help the model adapt to new data. At the same time, we will prepare training data by soliciting past weather and high tunnel climate data from other sources and collecting new data by setting up an on-site weather station. In addition, we will continue tuning the weather forecast DL model by exploring different input variables, neural network architecture, and longer period data sets. Objective 3: Understanding the socioeconomic aspects of greenhouse technology adaptation Our goal is to conduct 22 interviews by November 15th, 2023. A qualitative content analysis will follow with a comprehensive description of the investment decision journey, including the identification of discovery channels, sales approaches by technology and service firms that are more likely to culminate in transactions (as opposed to non-sale closings), marketing touchpoints with positive evaluations, and offer elements (factors) that enhance the probability of parties to establish long-term commercial relationships. In the second stage of the project, we pivot to the financial aspects of greenhouse technology adoption. Between January and June 2024, we will re-connect with study participants to construct enterprise budgets for representative operations and create automated spreadsheets for simulation and sensitivity analysis. This effort will include cash flow and feasibility analyses for alternative technologies being examined in the other project objectives. The final stage of objective 3 (intended to occur between July and December 2024) will evaluate the perceptions of CEA production companies' executives regarding alternative offering schemes. We will consider the results found in stages one and two to examine the executives' propensity to adopt a given technology when it is framed under alternative structures. Objective 4: Engaging our stakeholders through professional learning opportunities that contribute to workforce development The planned workforce development course will be offered in early 2024. Following the first course, we plan to expand its capacity to provide the long-needed professional education especially in the area of CEA engineering and technology. OSU reached out to a wide range of audiences through conferences and tours; but we plan to organize project-focused specific tours to discuss the technologies and outcomes, beginning in 2024.

Impacts
What was accomplished under these goals? Objective 1: Development of a data- and model-driven decision-making platform We have successfully integrated Koidra's interface software with the Priva climate control system at OSU. We have completed the selection of a wireless sensor system to be deployed at the three research sites (Arizona, NJ, and Ohio) and successfully tested the functionality of the wireless sensors using LoRaWAN gateways at OSU and UAz. The reading of these wireless sensors were integrated into Koidra's interface for further incorporation in the decision making platform that implements any necessary changes in the control set points. For the necessary AI framework, Koidra developed the foundational structure of a Digital Twin model. Koidra also introduced their new product Opera that will be integrated into the OSU climate controller to operate as an expert-driven decision support system. Opera enables growers or scientists to flexibly express their control logic through an intuitive, low-code mechanism on a web application. Koidra developed and continuously refined the AI autonomous control agent, based on the physics-informed Deep Learning and Model Predictive Controller framework. Through these tools and framework, Koidra can provide growers with two solutions that empowers them to manage their greenhouses effectively and efficiently: one approach allows for the expression of control logic through an intuitive, low-code platform, while the other approach introduces an AI-driven model designed for autonomous optimization of greenhouse conditions. In a concurrent effort, the Cornell team examined their AI integration into climate control and optimization, and made it ready for testing in research greenhouses. Mitigating energy demand was considered the major issue for lowering costs and reducing the carbon footprint. The project team expanded data processing and developed simulation models of CEA systems, including semi-closed greenhouses, plant factories with electric lighting, and integrated rooftop greenhouses. These models can predict and optimize indoor climates for crops under different climate conditions (such as AZ, FL, IL, NY, WI, WA, Iceland, Dubai, UAE, etc.). The AI-based control framework that combined physics-informed neural network and robust model predictive control demonstrated its adaptability to different climatic conditions for tomato cultivation in Ithaca, New York, and Tucson, Arizona. During simulations, the AI-based control framework reduced production costs by 46.4% and energy consumption by 57% when compared to conventional control methodologies. The findings from ten different global locations indicated the AI system's ability to curtail energy consumption. Objective 2: Validating the efficacy of new data- and model-driven decision making Using a Venlo-style "high-tech" greenhouse, OSU developed and examined a new leaf pruning strategy driven by PAR monitored at the lower leaves of plant canopy. Leaf pruning is a labor-intensive practice conducted conventionally weekly in commercial greenhouses with high-wire crop production. Approximately 60% of the planned first experiment was accomplished by the end of the first project year, generating data of spatial and temporal variations of light availability at the lower level of tall tomato crops, in addition to tomato crop growth, yield, and fruit quality. Preliminary data showed that the frequency of pruning and degree of leaf removal could be optimized based on the weekly integration of PAR measured near the bottom of the plant canopy. Using a "medium-tech" greenhouse with double layered polyethylene plastic films, UAz designed and implemented a three-zoned crop production system with autonomous drip irrigation to grow tomato crop in the project greenhouse. The project team initiated crop production with tomato and started collecting crop data, which will then be related to the environmental data as part of the model-driven decision making for the control settings. Toward examining implications for 'lower-tech' greenhouses or high-tunnels, OSU developed a prototype deep learning (DL) model for on-site weather forecasting to improve the site-specific solar radiation forecast from NOAA's HRRR (High-Resolution Rapid Refresh) forecasts. For data preparation, the project team harvested 6+ months of historical meteorological data (March and October in 2019, 2020, and 2021) from the HRRR data base, including actual time and one-hour forecasts, and from local weather stations in three different cities. Python programs were developed to automate the HRRR data, which includes 170 variables, to download and to extract 21 selected variables for the DL model development. Since weather stations have their data recorded in different formats, computer programs were developed for unit conversion, time synchronization, and handling of missing data. Three DL models were trained independently for Wooster, Ohio, West Lafayette, Indiana, and Geneva, New York. The training focused on seasonal change and high solar radiation periods (>400 W/m2) which has the most impact on the environmental conditions inside a high tunnel. The DL model development approach of combining HRRR actual time estimates with 1-hour forecast was shown to improve the forecast for the Wooster and Geneva locations, but using the HRRR 1-hour forecast alone was sufficient for Geneva. Through these analyses, the project team has identified the key variability of the most relevant HRRR variables for high tunnel climate prediction as solar radiation. Improved site-specific solar radiation forecast, compared to that of HRRR data, can lead to more reliable high tunnel climate forecasts for growers to take time-sensitive actions to prevent yield loss due to high temperatures. Objective 3: Understanding the socioeconomic aspects of greenhouse technology adaptation Objective 3 activities started during the first week of August 2023 (approximately 33 business days of activities to report during this first reporting period). Therefore, no major accomplishments can be reported at this time. Objective 4: Engaging our stakeholders through professional learning opportunities that contribute to workforce development A major accomplishment is the planning of ADVANCEA Academy's (https://www.ceaforum.org/advancea-academy) first offering of a 13-week online course titled 'Introduction to Greenhouse Environmental Control for Crop Production' that includes 26 selected lectures to be presented online during January-March, 2024. This effort involved professional greenhouse consultants from Delphy (The Netherlands) to provide six guest lectures as part of the course. ADVANCEA project members organized two short courses during the first project year: 1) 22nd UA-CEAC Annual Greenhouse Crop Production and Engineering Design Short Course, March 2023 at the University of Arizona, Tucson, AZ, and 2) 2023 Greenhouse Management Workshop, January, The Ohio State University, Wooster, OH. We also organized a symposium "2nd OHCEAC Annual Conference" focusing on CEA sciences and technology advancement, to deliver information relevant to our stakeholders. ADVANCEA outreach events also target the public audiences in addition to professionals. Each of the three research sites was visited by various national and international visitors, students, and groups of professionals. We achieved an agreement of collaboration with another SCRI SREP funded project team OptimIA (2019-2024) to collaboratively conduct monthly webinar series (Indoor Ag Science Café) beginning in August 2024. This webinar series has more than 1,300 subscribers whose focus is indoor vertical farming. With ADVANCEA participation, the focus will include greenhouses and engineering technologies, which will expand the target audience.

Publications

• Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Kim, C. & Lin, Y. (2023) Ohio State University (OSU) Columbus station report. NCERA-101 Annual Meeting, University of California, Davis, CA, United States.
• Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Kubota, C. (2023) Cultivating CEA innovation by integrating horticulture and engineering. Horticulture/Engineering Joint Seminar. Auburn University, Auburn, AL, United States.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Kacira, M. (2023) University of Arizona station report. NCERA-101 Annual Meeting, University of California, Davis, CA, United States.
• Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Ling, P. (2023) Ohio State University (OSU) Wooster station report. NCERA-101 Annual Meeting, University of California, Davis, CA, United States.
• Type: Conference Papers and Presentations Status: Other Year Published: 2022 Citation: Nguyen, N. M., Tran, H., Duong, M., Bui, H. & Tran, K. (2022) Differentiable Physics-based Greenhouse Simulation. Machine Learning and the Physical Sciences workshop, NeurIPS 2022, New Orleans, LA, United States.
• Type: Journal Articles Status: Submitted Year Published: 2023 Citation: Decardi-Nelson, B. & You, F. (2023) Harnessing AI to boost energy savings in plant factories for sustainable food production [Manuscript submitted for publication]. Nature Food.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Hu, G. & You, F. (2023) An AI framework integrating physics-informed neural network with predictive control for energy-efficient food production in the built environment. Applied Energy, 348, 121450. https://doi.org/10.1016/j.apenergy.2023.121450
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Both, A. J. (2023) Rutgers University station report. NCERA-101 Annual Meeting, University of California, Davis, CA, United States.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Decardi-Nelson, B. & You, F. (2023) Optimal energy management in greenhouses using distributed hybrid DRL-MPC framework. 33rd European Symposium on Computer Aided Process Engineering, 52, 1661-1666. https://doi.org/10.1016/B978-0-443-15274-0.50264-X
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Hu, G. & You, F. (2023) Energy management for controlled environment agriculture based on physics informed neural networks and adaptive linearization based data-driven robust model predictive control with AI. 33rd European Symposium on Computer Aided Process Engineering, 52, 2205-2210. https://doi.org/10.1016/B978-0-443-15274-0.50351-6
• Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Kacira, M. (2023) Advancement of Plant Sensing Technology for Sustainable Crop Production Under Controlled Environment. OHCEAC Annual CEA Conference, Ohio State University, Columbus, OH, United States.
• Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Kacira, M. (2023) Monitoring Greenhouse Environments. 22nd UA-CEAC Annual Greenhouse Crop Production and Engineering Design Short Course, The University of Arizona, Tucson, AZ, United States.
• Type: Conference Papers and Presentations Status: Other Year Published: 2023 Citation: Kacira, M. (2023) Resource use Efficient and Precision-Controlled Environment Agriculture. Cornell University Ezra Round Table Seminar Series. Online.