Source: ACREAGE FARMS LLC submitted to NRP
SCALABLE CROP MONITORING SYSTEM FOR TRAVERSING VERTICAL FARMS AUTONOMOUSLY
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
SMALL BUSINESS GRANT
Reporting Frequency
Annual
Accession No.
1031915
Grant No.
2024-33530-42005
Cumulative Award Amt.
$174,980.00
Proposal No.
2024-00347
Multistate No.
(N/A)
Project Start Date
Jul 1, 2024
Project End Date
Feb 28, 2025
Grant Year
2024
Program Code
[8.13]- Plant Production and Protection-Engineering
Recipient Organization
ACREAGE FARMS LLC
1511 AVIATION CENTER PKWY STE 202
DAYTONA BEACH,FL 32114
Performing Department
(N/A)
Non Technical Summary
The US has seen an increase in population in recent years and a decrease in domestic food production due to climate change and extremely high resource requirements. This has resulted in food shortages, a higher demand for food, and greater food insecurity across the US. In recent years, new farms are being constructed as high-density, vertical farms, as opposed to traditional field farms to vastly multiply the space utilization efficiency. However, this relatively new industry struggles to be sustainable due to excessive start-up and operational expenses. In particular, monitoring crops in these huge facilities is time-consuming for workers and requires companies to buy thousands of statically-placed sensors to ensure the farms are operating as efficiently as possible.The main focus for this work aims to alleviate these issues by developing an autonomous and mobile crop monitoring system built to survey vertical farms. The system will be able to move in three dimensions over the crops while also traversing multiple tiers with a single set of environmental sensors and cameras. Measuring the 3D environmental parameters will provide insights for optimal growth at individual crop sites. Additionally, using visible light and NIR cameras, plant nutrient deficiencies can be identified days in advance of a worker spotting the issue, preventing crop loss. Two variations of the system architecture will be studied to provide customers with an option between higher utility/ actionable data and a lower cost version. Utilizing this technology will allow farms to increase yields and free up their staff to complete other necessary tasks. Another benefit for this solution is that it can run continuously and autonomously while providing warnings to growers or API-based commands that could run necessary processes (e.g., watering). The proposed system can significantly reduce a vertical farm's startup cost by removing thousands of components normally used for crop monitoring. The main objectives for Phase I are 1) Build and integrate two 3D monitoring systems, 2) Collect environment data, provide data analytics, and control system commands, and 3) Complete an economic assessment of commercial viability. Completing this work will allow farms to properly optimize the growth environment which can lead to increased energy efficiency while also increasing consistent and stable food production which is becoming more urgent as time passes.
Animal Health Component
20%
Research Effort Categories
Basic
10%
Applied
20%
Developmental
70%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
20572102020100%
Knowledge Area
205 - Plant Management Systems;

Subject Of Investigation
7210 - Remote sensing equipment and technology;

Field Of Science
2020 - Engineering;
Goals / Objectives
The overall goal for this program is to create an autonomous monitoring system prototype that is on a mobilized apparatus to traverse across multiple tiers of plant sites while recording environmental and plant health parameters within a small scale testbed. We aim to demonstrate the ability to process the collected data to provide insight on plant health indicators and help farmers make informed decisions on how to increase yields. To complete this work, a lab scale, 4-tier, hydroponic grow space will be built to collect real time data using the monitoring system. The feasibility of the mobile monitoring system will be assessed through two variations consisting of differing sensors. The resulting information from Phase I will be two potential offerings which can collect the plant growth and environmental parameters in 3D space throughout the farm.Through Phase II, the systems can be incorporated into a local vertical farm to then study integration challenges as well as further demonstrate functionality. Additionally, this phase would incorporate additional sensor components that would otherwise be a financial struggle (e.g., multispectral imaging). At this point, all data collected thus far will be processed through machine learning algorithms to provide actionable insights to enhance the growth environment and increase crop productivity.The proposed research is predicated on the following research questions:How can a mobilized system be designed to function as a plug and play solution for other farms without intruding on current designs?What camera-based indicators can be used to assess plant characteristics (e.g., plant size, color, disease, and pests)?What are the most effective combinations of sensors to provide highly accurate data collection while at an affordable cost?What is the most effective way to display the collected data in a human interpretable way?What are the projected economics of manufacturing, distributing, and installing the monitoring system?Furthermore, pending the availability of time, additional exploratory questions will be asked as the platform is in development:How can ML and AI best be used to provide useful data analysis and operational recommendations?The following objectives will be necessary to answer the research questions:A) Operate a 4-tier, automated, hydroponic testbed.B) Build and integrate two 3D monitoring systems within 3 months of project start.C) Collect environment data, provide data analytics, and control system commands.D) Complete an economic assessment of commercial viability.
Project Methods
The project is split into nine tasks with a tenth prepared, if time is available during the Phase I otherwise it will occur in Phase II. Each of the tasks aim to complete a portion of each of the four objectives and have the time period when the task will be completed. The tasks are as follows:A) Operate a 4-tier, automated, hydroponic testbed.B) Build and integrate two 3D monitoring systems within 3 months of project start.C) Collect environment data, provide data analytics, and control system commands.D) Complete an economic assessment of commercial viability.Task 1 (Objective B) (Month 1): Determine Sensor Variability and ReliabilityFor the project to begin, the various sensors being used need to be coded to test and verify their functionality. Different sensors will have different levels of performance and/or fidelity and, when creating sensor suites with varying levels of performance due to interchangeable sensors, we must identify the baseline differences between all sets of sensors. There will be two primary tested systems built around different edge processing devices: a high fidelity version with an NVIDIA Jetson Orin Nano and a low fidelity option using a Raspberry Pi 4.Task 2 (Objective B) (Month 1-2): Wiring a Sensor Suite & PCB CreationDifferent combinations of sensors will be experimented with to view the optimum pairing of cost vs. sensor data fidelity. The collection of sensors will be able to log measurements for temperature (°C), humidity (%), light intensity (μmol s-1 m-2), airspeed (m/s), and imagery of the plants in visible and near infrared light. Once the sensors are all finalized and functioning, custom PCBs will be created in Eagle PCB, ordered, assembled in house, and encased in a 3D printed housing.Task 3 (Objective A) (Months 1-2): Construction of the TestbedA four tier testbed will be built to represent the layering that occurs in commercial VFs in our facilities at the ERAU Micaplex. The testbed will be at a lab scale and will have approximate dimensions of 36' x 72' x 96'.Task 4 (Objective B+C) (Months 2-4): Developing a Mobilized Platform Over the Plant CanopyWith the testbed built, a solution to maneuver the monitoring system across all 4 tiers of the plant canopy will be devised. This design will provide access to the entire plant growth area at intervals that will be flexible based on plant spacing. Aluminum extrusion material will be used initially, but may need to be later changed to a material better suited for commercialization. The system will have onboard battery packs operating at 12V and 5V to power the unit and then dock at a homebase position to recharge, doing so will reduce wiring necessary.Task 5 (Objective C) (Month 3-4): Develop Autonomous Movement and Data Collection SoftwareThis task will develop the software solution to autonomously move the sensor suite with the mobile platform and collect data using the sensors. The autonomous capabilities of the sensor suite will increase productivity by collecting data at regular intervals across the growing environment. The solution will traverse the growth environment and collect data with minimal amounts of input from the user by placing them on-the-loop rather than in-the-loop.Task 6 (Objective A+C) (Months 4-8): Create 3D Environmental Distribution Dataset using MMSThe monitoring system will be placed on the mobile platform and take sensor readings over 3D space. The MMS will autonomously move around the testbed and collect data from the sensors at specific points and write it to a database. This dataset will provide insight into the system's ability to collect data autonomously and create an environmental map of the testbed.Task 7 (Objective C) (Months 4-5): Create Data Analytics Dashboard for Environmental MonitoringThis task will create an interactive dashboard for users to visualize the collected data. The dashboard will have a variety of graphs and other displays to show the collected data from the database. The user will have the ability to select which parameters they would like to view, filter data by the date it was collected, and filter images by the location they were collected in the test bed. These baseline functions will provide baseline insights about the test bed's 3D environment.Task 8 (Objective C) (Months 6-8): Develop Data Analytics Toolkit for DashboardTo further the functionality of the dashboard, a data analytics toolkit will be created. The toolkit will provide automated, advanced features to the dashboard that presents deeper analysis of trends found in the data. These trends will be displayed to the user on the dashboard in a graphical, human interpretable format. The data analytics toolkit will provide valuable information to growers about non-optimal conditions that are being experienced in the growth environment.Task 9 (Objective D) (Months 7-8): Technological and Economical Assessment of the Mobile Monitoring SystemAll of the data and equipment developed over the Phase I period will be evaluated and an in-depth economic feasibility of the MMS will be documented for commercialization. The materials used for the prototype will be assessed to determine better alternatives for manufacturing. The findings from this work will be presented in the close out document for the Phase I program as well as form the basis for the Phase II proposal. The team plans to attend Indoor AgCon in 2025 to showcase the developed work and attract potential clients.Pending Availability - Task 10 (Objective C): Machine Learning & Localization of Plants using AI/MLIn the case that the project is completed at a faster rate than anticipated, additional research can be accomplished on the implementation of useful AI/ML to provide useful data analytics and make recommendations. The first step in this process is determining if current computer vision models can perform the localization of all images of our plants in the environmental dataset. Plant localization will provide foundational functionality that is necessary to perform image based plant health monitoring. Plant localization will identify pixels of an image containing a crop, allowing for only relevant portions of the image to be used when identifying nutritional deficiencies that may be present. The model will output binary masks that will be overlayed on the original image in the dashboard.

Progress 07/01/24 to 02/28/25

Outputs
Target Audience:The customers that would be utilizing this monitoring platform can be divided between three different vertical farm (VF) sizes: small (converted shipping containers), medium (between four to six tiers in a warehouse-style setting), and large scale VFs (massive complexes with greater than six tiers). After expanding customer discovery efforts, the intended audience for the initial MMS offering would be medium to larger scale VFs as these farms have the most crop coverage needed. Reaching these clients will occur both directly to currently operating VFs as well as farm manufacturers. By working with farms under construction, we can immediately integrate this innovative monitoring system in their farms while greatly reducing their crop monitoring investment costs. Through integrating the Mobile Monitoring System (MMS) into these facilities, it can instantly provide workers with 24/7 monitoring activity and recommend specific tasks to be completed or problem areas to address (e.g., pest removal or nutrient adjustments). Additionally, because of the ease of scalability and complete farm coverage, workers can remotely access the status of crops in difficult to see areas which is especially useful for farms above five tiers. Through the proof of concept being created as part of the current Phase I work, Acreage Farms has been developing relationships with local Florida farms and universities for future collaborations as part of the Phase II process. In particular, the team is currently communicating with the University of South Florida to explore obtaining an office space to carry out various growth studies with different stress conditions imposed on the crops to develop image datasets for the vertical farm environment. These datasets will be crucial for advancing our future machine learning algorithms and provide a better product to our intended clients. Changes/Problems:Two major changes have occurred during this work. First was the substitution for a Jetson Orin Nano for a Raspberry Pi Zero 2W. The reason for this change was due to the Jetson's physical size being too large for this application and it having many compatibility issues with various environmental and imaging sensors. The change in single board computers enabled the project to have a smaller sensor suite option which increased the potential for the technology to function as intended. The second major change was the use of cable carriers rather than battery power with wireless charging. The reason for this change was to allow for the continuous and dedicated development of the movement system without the additional complexities and challenges. The inclusion of batteries and charging capabilities will take place once the movement system is completed and reliable. Additionally, a secondary prototype station will first have this added benefit before converting the current testbed, allowing for a smooth transition of development. What opportunities for training and professional development has the project provided?Thus far, the project has provided numerous opportunities for both training and professional development. In particular, team members have had to cross-learn many skills such as different software programs (e.g., Python and Next.js) to operate and maneuver the sensor suite across the crops. Other skill development has been related to hardware integration, circuity, and computer aided design. Additionally, an undergraduate student intern was hired to learn about vertical farming and plant science while managing the crops growing throughout this work. Beyond these, this project has provided the team the ability to expand their business network and growth opportunities by connecting with local CEA farms and universities for future collaborations. How have the results been disseminated to communities of interest?Current results have not been released to communities of interest. The team will be attending Indoor AgCon in March '25 to complete further customer discovery as well as discuss our progress to gain interest from potential clients. What do you plan to do during the next reporting period to accomplish the goals?To complete the goals set out at the beginning of this project, the team will be focusing on two main directions. First, the program currently operating each sensor system needs to be upgraded to autonomously move throughout a tier, exit, move to the next tier, enter, and complete a scan. With this change in operation, the system will achieve the main goal of collecting environmental and crop image data using a single sensor suite over multiple tiers. While this data is being recorded, the analytics dashboard needs to intuitively display the results for a user. Once these two directions are accomplished, demos can be shown to previous interviewees who were interested in seeing the team's progress. Doing so will help improve the core technology for its intended users and can aid in determining economic and technical viability. Beyond these main directions, the commercial feasibility of the technology will be further assessed through market research, exploring pricing options, and customer discovery.

Impacts
What was accomplished under these goals? The goals for this project are split between four overall objectives. First, the hydroponic testbed was constructed to grow four tiers of 45 crops. To account for various vertical farming environments, three LED lighting options were included (i.e., white, pinkish, and blue-red) along with mylar layers placed beneath each tier's hydroponic channels to prevent the lights from bleeding through into the next tier. This testbest also had fans on each tier installed and a central water tank for water to circulate out and then drain back. A separate grow tent was utilized to house a seed starting station which allows the seeds to germinate and grow for two weeks before being transferred to the testbed. With all of the systems in place to grow crops, multiple cycles have been completed from seed-to-harvest which has been donated to nonprofits, friends, and family. Continuous improvements are being incorporated to our growing methodology (e.g., water condition tracking, cleaning, substrate usage). With the vertical farm testbed operational, the project's core technology (the mobile monitoring system) was installed on each of the largest sides to develop a higher and lower fidelity data capture solution. This gantry-style system enables 3D motion of a sensor suite taking environmental and crop image data. Two electronics bays were wired on each side to transfer power as the system moves around. To mitigate slippage from the stepper motor enabling vertical motion, a counter weight was added which equalizes the force on each side of the motor. Additionally, a hinge and a linear actuator were used to allow the sensor suite to enter or exit from a tier. Concurrently, the next objective required each sensor suite to be soldered together, encompassing a collection of environmental (i.e., temperature, relative humidity, CO2 concentration, wind speed, and light intensity) and imaging sensors, and inserted into custom 3D printed cases. The two sensor systems varied from one another due to higher quality environmental sensors and the inclusion of a thermal camera in the higher fidelity model. Moving the system around was made possible through communicating between single-board computers to microcontrollers via MQTT. Data collected is transferred to a network attached storage device which will later be used for training machine learning models. To visualize the measurements, a web-based dashboard is currently in development. At the midpoint of this work, initial renders have been created to demonstrate concepts of the application and a template has been programmed consisting of a functional navigation bar, user authentication, and is mobile friendly. The last objective for this work is based on determining the economic viability for this technology. One method for exploring this possibility has been through customer discovery to meet and learn from those working in the controlled environment agriculture field. Additionally, a modular bill of materials is being created to explore system costs based on various sizes of vertical farms. Another aspect to improve the viability of this work is a commercialization plan which is being drafted and then reviewed by an expert in the crop monitoring space

Publications