Source: LATERAL.SYSTEMS LLC submitted to NRP
LATERAL.SYSTEMS AQUAPONICS EDGE PLATFORM
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
SMALL BUSINESS GRANT
Reporting Frequency
Annual
Accession No.
1028617
Grant No.
2022-33530-37411
Cumulative Award Amt.
$181,650.00
Proposal No.
2022-01383
Multistate No.
(N/A)
Project Start Date
Jun 1, 2022
Project End Date
Jan 31, 2023
Grant Year
2022
Program Code
[8.13]- Plant Production and Protection-Engineering
Recipient Organization
LATERAL.SYSTEMS LLC
3121 S MOODY AVE
PORTLAND,OR 972394505
Performing Department
(N/A)
Non Technical Summary
Lateral.Systems LLC is an Agricultural Technology startup company. Our mission is to enable generations of farmersby accelerating the intersection of promising modern technologiesto advance sustainable production of food. Our primary focus is developing a compute platform for indoor agriculture. A platform consists of hardware, data storage devices, software applications, and networks that support the processing and exchange of electronic information. In this case, we are developing a platform used for monitoring and maintaining biological, chemical, and physical elements within indoor growing environments such as greenhouses, repurposed warehouses, dry docks, manufacturing facilities, barns, and shipping containers. Our initial target with this industry is aquaponics, one of the fastest-growing industries in the U.S. (Mordor Intelligence, 2021).A key motivation for making aquaponics our beachhead is the global water crisis. The World Bank (2021) estimates that future demand on fresh water resources by all sectors will require as much as 25-40% of water to be reallocated from lower to higher productivity, particularly in water stressed regions. Given that agriculture accounts for approximately 70% of freshwater consumptive withdrawals globally (Goddek, et al., 2019), supporting the growth and success of aquaponics represents a highly promising approach to sustainably feeding the world. Aquaponics is organic agriculture that pairs fish and plant cultivation into a mutually beneficial growing system using an average of 5% of the water typically used to grow the same plants in soil (Kelly, 2021). Well-designed aquaponics produces an average of 30-40% higher yields than plants grown in soil using a fraction of the energy and other costly external inputs (Fernadez-Cabanas, et al., 2020) while producing little pollution (Mallick et al., 2021).As promising as this novel farming approach is, aquaponics comes with risks and requires deep knowledge to be successful (Turnšek et al., 2019). However, a worldwide survey of commercial aquaponic growers found that only 59% of respondents had some relevant prior knowledge during the start-up phase; 41% claimed to have insufficient knowledge of both fish and plants in their first year. Approximately 1/3 of aquaponic businesses are started by entrepreneurs with no prior training or experience in growing fish or plants (Greenfield, 2020). High initial financial outlays on aquaponics systems and a steep learning curve pressured by tight profit margins often result in failure or at best, an average of 7-to-12 years to pay off investments (Baganz et al., 2020). Our smart indoor farm monitoring and management tools will advance adoption and success of aquaponics farming by significantly reducing pain points and risks.Existing indoor agricultural technologies such as digital probes for collecting data suffer from an inability to compile and interpret information due to a number of technical challenges. Currently, a non-proprietary modular plug and play platform approach tuned for indoor agriculture to integrate relevant sensors and the data they produce is non-existent. The fusion of relevant and disparate datasets is required to aide farmers to improve their respective growing operations (Open Ag Data Alliance, 2021). Our technology will advance USDA strategic goal 2 (maximizing the ability of American agricultural producers to prosper) and NIFA priorities by accelerating controlled agricultural growing systems by enabling precision controls for indoor growing that conserve water and energy while producing healthy food for growing populations.The goal of this USDA NIFA SBIR Phase I feasibility study is to advance critical enabling features of an on-premise smart service for farming applications. The scope of this project is to test and evaluate a well-established open-source compute framework to determine if this software and hardware technical approach is an effective solution to commercialize our product and services. This open-source software framework is designed to facilitate interoperability between multiple devices (such as sensors) and software applications.We will know that this particular framework is a feasible solution if we are able to validate that we can use the software stack to activate data flow (i.e., movement of data from various sensors in the fish tanks and plant beds to software), data fusion (e.g., overlaying multiple pH and temperature readings over time), and control logic (i.e., a key part of a software program that controls the operations of the program. The control logic responds to commands from the user and it also acts on its own to perform automated tasks structured into the program).The development and testing steps are crucial to ensure that we select a solid software stack that will enable developers to code better and faster with a well-documented codebase. Testing to see how easy it is to run tests on the platform now will helps us to make informed decisions about how easy it will be to maintain the system, detect and fix common bugs or performance issues and tweak features without eating up too much time. Determining if the platform has a well-defined scalability matrix that works well for vertical scalability (adding features) and horizontal scenarios (handling increased volume of users and transactions on the platform) while maintaining security both on the client and server side will ensure that we avoid choosing the wrong stack.If the open framework we test proves to be suitable, this will advance our development towards beta testing our product and services within one year in preparation of a soft launch shortly thereafter. Revenue generated from the soft launch will fund further development of cloud-based features that will enable more advanced data analysis, data modeling, and predictive analytics that have the potential to support farmers' long term business management and increase their productivity and profits. Choosing the wrong stack could lead to additional time to adapt and rebuild from the ground up or potentially create extra work down the road with tasks such as tracking frequent update cycles that require constant changes to keep software applications running with the latest codebase, which may become unsustainable.
Animal Health Component
75%
Research Effort Categories
Basic
(N/A)
Applied
75%
Developmental
25%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
20553102020100%
Knowledge Area
205 - Plant Management Systems;

Subject Of Investigation
5310 - Machinery and equipment;

Field Of Science
2020 - Engineering;
Goals / Objectives
We are utilizing an open-source approach as a base for future Agricultural Technology (AgTech) innovations to enable an on-premise smart edge service for farming applications. The purpose of this research is to accelerate our software prototype design by advancing critical enabling features of our conceptual platform to the pilot phase. This Phase I feasibility study involves selecting an open-source "stack" (operating system, web server, database and program language) and designing a study to identify specific processes and functions that must be monitored to drive the development of the rules engine and eventually enabling predictive analytics functions. The research question is: What is the feasibility of extending an open-source software approach for productization of our "on-premise" edge solution that does not require internet services to bridge the digital divide for rural and urban farmers?Goal 1: Determine the feasibility of EdgeX as an effective approach for our product.Objective 1a: Leverage a standardized open interface to enable multiples device types to "bolt in to" the platform to achieve sensor fusion. Raw sensor data are essentially useless until organized for analysis. As is the case with humans that combine data points from multiple senses to determine if a substance is edible or rotten, data fusion involves sensor data integration followed by reduction into compiled datasets that can be classified and used as actionable data. The application program interface (API) for each sensor is different and different types of sensors are wired in various ways such as GPIO and I2C configurations. Thus, we will form a hybrid baseline sense of sensors to test with the goal of achieving data flow and data fusion.Objective 2a: Define the control logic in preparation to demonstrate controller functions.Objective 3a: Demonstrate the data flow of sensor data into the core services of the stack in preparation for applying control logic. To do this, we will program the device services for the stack. This involves programming the device profiles for a suite of baseline sensors and creating the device registry. The devices are sensor processing units (i.e., pH, dissolved oxygen, temperature); a device profile describes the set of attributes (services and/or features) and the value properties associated with a particular of sensor (i.e., the type of data the sensor collects like pH, temperature, etc.). A device registry is a container of devices with shared properties used to manage, monitor, and maintain registered device types. Each device managed by a device service has an association with a device profile, which defines that device type in terms of the operations it supports (i.e., in this case, the measurement it performs).Objective 4a: Determine if EdgeX will suit our needs.Goal 2: Lay the foundation for our Private Cloud design.Our second goal is to lay the foundation for moving our pilot design to commercialization by designing the architecture of the Private Cloud relational database. Data must be archived in a relational database to facilitate pulling useful data. In phase I, we will design this architecture and a decision tree to program the rules engine. This involves establishing a plan for which parameters will be included in the baseline sensor array model and creating tables that depict the range of tolerances for plants and fish typically grown in aquaponics farms. This table is necessary to create a decision-making tree (If/Then/ Else) for our control logic, a necessary first step for programming the rules engine that applies rules to make predictions, diagnose problems, and trigger actions.For instance, nutrient management is necessary to counteract imbalances and provide an optimum nutrient solution needed to maximize yield and prevent physiological plant disorders (Suhl et al., 2016). If any of the nutrient measurements approach the bottom or the top of the tolerance range for a specific crop, then the rules engine will interpret these readings to trigger the dashboard alert or alarm to warn the farmer that some action is needed before the crop ir fusg are stressed. Depending on the nature of the impending imbalance, the rules engine may trigger an injector to add the appropriate buffering regime (Rakocy et al., 2006) such as providing nutrient supplementation (e.g., adding chelated nutrient forms), (Roosta & Hamidpour, 2011). These triggers will occur at the edge (i.e., on-premises) rather than relying on internet connectivity to function. Objective 1b: Design a study to identify specific processes and functions that must be monitored to drive the development of the rules engine database, a foundation of the Private Cloud. We assume situating advanced analytics in the cloud may be more efficient to build more complex data models and enable sharing of data and data models with researchers and between farmers. Sending every data point to the cloud would result in prohibitively expensive cloud service fees, thus, we must strategically define the purpose and objectives of our relational database to reduce data at the edge and send only necessary data to a cloud for advanced analytics.
Project Methods
Before we can determine the feasibility of EdgeX as an effective approach for our product and services the following efforts will be made to enable data capture and data flow:Develop open interfaces for a hybrid suite of sensors to test the data flow and ensure modularity, integrability and interchangeability of the individual components.Develop pseudocode control logic (i.e., descriptions written for humans to understand, not machines).Defining the control logic is core to the intelligence of our growing decision subsystem in terms of machine learning and artificial intelligence preparedness. So, we will model design features overview and then fill in each part that requires more detail with pseudocodes to provide a high level description of the operating principles of our microservices.Write a device profile specifying the data exchange and communications with a sensor/device.Validate the device profile (i.e., check that code runs correctly and fix errors),The next effort is to get an EdgeX instance up and running - with the device service for the particular sensor/device. An EdgeX instance functions as a server running microservices. Our instances will have templates we will use to clone new instances with the existing configuration we establish, saving quite a lot of time in the long run. This effort will involve:Providing device profiles to device services, provision the devices, send data through EdgeX, and capture data sent through EdgeX (needed in the application),Describing the data (i.e., create a schema) for the data sent through EdgeX.Providing visualization of the data as a test of being able to digest data into the software EdgeX stack.The next effort will be to perform a gap analysis to evaluate if EdgeX is robust enough to provide agricultural services or if we should use a hardened platform (such as LAMP), which would require a larger workforce of software programmers and engineers and a longer, more involved development process. This gap analysis effort will involve designing a comparative analysis of precision agriculture cloud services to examine current usage models to work towards defining the purpose and objectives of our relational database and setting a baseline for our AI analytic services.Another effort is to design a full research plan for a study to collect empirical evidence of data requirements and computational analysis functions necessary to generate useful data models in a cost-effective manner. This effort will involve conducting a comparative analysis of precision agriculture cloud services to examine current usage models to work towards defining the purpose and objectives of our relational database and setting a baseline for our AI analytic services. Designing the study will include articulating a list of entities and a list of attributes to validate, a plan models the tables and fields in the database, and a plan for how to establish table relationships and business rules. As part of this last effort, we will also continue adding to the decision-making trees to be used to program the rules engine.

Progress 06/01/22 to 01/31/23

Outputs
Target Audience:Commercial Indoor Growers. Our primary target customers are indoor commercial farmers interested in growing year-round specialty crops, including both full-time and part-time growers. Specifically, we anticipate targeting farmers who grow non-commodity, specialty food crops (i.e., lettuce, spinach, and arugula, microgreens such as watercress and, tomatoes), medicinal, and culinary herbs. Our initial targets are indoor food producers in the U.S. starting off with growers in the Pacific NW region close to where we are based and then expanding nationwide. In approximately five-to-seven years, once we have a solid foothold in the domestic market, we plan to expand to an international market with particular focus on regions experiencing fresh water scarcity coupled with access to inexpensive energy resources, two important drivers to the expansion of the global indoor farming market. As part of the TABA funded effort we interviewed current farmers, outdoor farmers exploring the possibility of expanding operations to include indoor growing, GAP certification consultants, inspectors, farm extension agents and indoor ag consultants, ag equipment distributors, companies offering complementary technologies, fish hatchery managers, ag researchers, and science educators. We learned nuances related to current monitoring practices, desired capabilities, and market potential in the U.S. and abroad. Based on these data, we are working towards expanding the application of our monitoring technology to include hydroponics and plant germination rooms in addition to our initial target market, aquaponics growers. Researchers. Our secondary target market are researchers involved in medical research involving plants and aquatic organisms and those involved in agricultural research. Medical research community. We envision that our product will have a smaller, though significant, market amongst universities and private company researchers involved in studies related to cultivating food as medicine for disease prevention and renewing health by matching nutritional profiles to suit patients' specific genetic profiles. We also envision a market amongst government researchers such as: NASA research for space colonization; NOAA research related to reducing pressures on wild fish stocks; FEMA research for emergency food provisioning; and researchers involved in military readiness work related to cultivation of food to ensure steady provisioning of nutritious meals in insecure locations such as regions experiencing civil unrest and/or nature disasters. Agriculture research community. We envision a market amongst the agricultural research community involved in: Developing complementary agriculture technology for indoor growing, Research focused on fish diseases and health, Research focused on the effects of lighting on plant growth and nutrition, Development of data dictionaries for machine learning training to enable AI. Researcher working on creating grow algorithms for particular crops to: reduce external input requirements, increase or decrease particular nutrients or mineral content. Educators. Another secondary market are members of the educational community involved in secondary and post-secondary level science education, science museums, and those involved in prison rehabilitation job skills training. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Although the primary goal of our of technology is to provide data insights to support healthy, bountiful crop production, the mediation options and message notifications has educational value for users. We anticipate that beginner farmers with less than 10 years of experience (as defined by the USDA) will draw on the graphic data that depict and forecast trends and the written mediation options will support increased understanding of the biochemical relationships overtime. We also anticipate that experienced farmers will benefit from data depictions to share their rationale for management protocols when the system is in balance and for treatment choices when one or more variable exceeds a tolerance range of pathogens infest the growing systems. Our goal is not to create utter dependancies on the technology, but rather to increase growers' awareness of the inter related dynamics found within the growing environment such that they are able to improve on their management protocols in response to precise and accurate data signals. How have the results been disseminated to communities of interest?Because Phase I was a feasibility study, dissemination of our results is largely limited to the EdgeX Foundry Community and people who are affilitated with member companies such as Intel Corporation and IoTech Systems. We jointly fashioned press releases and website postings with Intel, IoTech, and Open Commons and posted updates on our own website. We have also been interviewed by local publications for inclusion in articles focused on agricultural technology innovations, such as the Oregon based Strategic Economic Development Corporation's quarterly journal that highlights innovations in Yamhill, Polk, and Marion counties. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? We completed NIFA SBIR Phase I goals and objectives Goal 1: Determine the feasibility of EdgeX We successfully determined that an AgTech platform based on EdgeX, an open-source edge platform under the Linux Foundation Edge Foundry umbrella, provides a robust, feasible solution for our monitoring platform. We validated sensor data fusion with data flow from multiple water and ambient sensors utilizing control logic while visualizing live sensor data using this framework. We validated that leveraging EdgeX's modular and robust open interfaces for rules and analytics engines will enable solutions LATERAL is planning to deliver. Objective 1 - Completed: Leverage a standardized open interface to enable multiple device types to "bolt in to" the platform to achieve sensor fusion. Method 1: Develop open interfaces for this hybrid suite of sensors to test the data flow and ensure modularity, integrability and interchangeability of the individual components. We successfully evaluated and exercised the open device service interfaces of EdgeX by testing sensor dataflow utilizing standard MQTT protocol delivering payloads to the EdgeX core data framework then feeding into the Grafana visualization engine, demonstrating an end-to-end data solution delivery. Objective 2 - Completed: Define the control logic Method 1: Develop pseudocode control logic We developed an initial set of sensor trigger points based on safe operational and optimal growing ranges for 3 species, forming the basis of a catalog of custom grow profiles, to architect pseudocode control logic that will feed the rules engine to handle notifications. Objective 3 - Completed: Demonstrate the data flow of sensor data into the core services of the stack in preparation for applying control logic. Method 1: Write a device profile specifying the data exchange and communications with a sensor/device. We developed device profiles for multi-sensor array device suites sending sensor data to a custom "LATERAL" EdgeX device profile service that recognizes unique devices with unique identifications and grouped sensor values with accurate time stamps delivered to a time series database (InfluxDB). Method 2: Validate the device profile We validated that sensor data is being captured correctly by directly monitoring the MQTT broker for incoming data payloads and comparing the fused data handled by EdgeX. The EdgeX data forwarded to our Grafana visualization service was validated to represent accurate data and timing precision providing a high level of confidence in EdgeX's ability to handle data flow. Method 3: Get an EdgeX instance up and running - with the device service for the particular sensor/device. The basic device service prototype is functioning as expected. We evaluated EdgeX on a series of headless Ubuntu Servers (22.04.1 LTS) utilizing Intel Core i3 and i7 NUC platforms running industry standard Linux Docker containers. We developed Docker Compose files to build, start and stop the EdgeX operational stack in a modular way to replicate on our various testbeds. Method 4. Provide device profiles to device services, provision the devices, send data through EdgeX, and capture data sent through EdgeX (needed in the application). We validated that our wireless edge is integrated, sending data flow through variousapplication layers. The prototype LATERAL device service is functioning as expected processing MQTT message payloads to the EdgeX data pipeline with complete data flow from multiple microcontrollers wirelessly delivering various sensor data to the LATERAL deviceservice forwarded to a Real-Time-Sensitive database (InfluxDB) relayed to a visualization engine. Method 5: Describe the data (i.e., create a schema) sent through EdgeX. We experimented using a MQTT Payload approach to sending data and created an initial schema for data sent through the EdgeX framework putting us in a solid position to explore other processor-based solutions (i.e., Intel-based x86 and ARM) to address potential long term supply chain concerns. Method 6: Provide visualization of the data as a test of being able to digest data into the software EdgeX stack. We prototyped a user interface to integrate visualization dashboards fed by the EdgeX framework. Again, we successfully tested end-to-end data flow into a Grafana visualization engine demonstrating sensor visualization integration with EdgeX sensor data with a popular TIG stack (Telgraf, InfluxDB, and Grafana) time series data visualization engine. Objective 4 - Completed: Determine if EdgeX will suit our needs. Method 1: Perform a gap analysis to decide if EdgeX is robust enough for our ag. services or if we should use a hardened platform (such as LAMP)... We completed a gap analysis and validated that EdgeX will suit our needs. EdgeX Foundry has proven itself mature with industry leader support. Goal 2: Lay the foundation for our Private Cloud design. Objective 1 - Completed: Design a study to identify specific processes and functions that must be monitored to drive the development of the rules engine database... Method 1: Design a comparative analysis of precision agriculture cloud services... We designed and conducted a comparative analysis of precision agriculture cloud services to determine desired and priority functionality to provide at the edge and in the cloud where advanced analytics will take place. We conducted extensive interviews, attended lectures at professional conferences, and continued our literature review to validate trigger mechanisms that cause cascading biochemical responses to add to the decision-making tree. We made progress on defining our initial version and subsequent version release timeline, which informed setting the baseline for our AI analytics services involving trends analysis and predictive analytics. Thus, we made progress defining the purpose and objectives of our relational database and began programming the rules engine using the EdgeX Rule Engine development package. Method 2: Create a full research plan for a study to collect empirical evidence of data requirements and computational analysis functions necessary to generate useful data models at the cloud level in a cost-effective manner. We developed a research plan to test multiple database instances on the multiple EdgeX platforms fed from various sensor sources. Various sensor suites were utilized to ensure that data can flow through different EdgeX instances on different platforms and then fuse together at the visualization service level.We experimented with a single instance of Grafana connected to multiple remote time-series InfluxDB databases being served on various architectures on the latest EdgeX release. We successfully tested our edge framework dataflow on a number of Intel IA64 and ARM64 architectures and were able to visualize sensor data on a real-time Grafana integrated dashboard with the entire software stack running on the edge platform. Method 3: Continue adding to the decision-making trees to be used to program the rules engine. We documented evidence-based operational and optimal tolerance ranges to establish grow profiles to feed our control logic to guide preliminary development of our control logic, a necessary first step to inform programming the rules engine that applies "rules" to make predictions, diagnose problems, and trigger actions. We reviewed EdgeX capabilities that will support decision making capabilities to predict and handle system level messaging. EdgeX supports various rule engines (eKuiper to include comparison with gRules) and notification capabilities.We are satisfied that this capability will be well suited for decision making that will support machine learning and predictive analytics addressing water quality and air quality conditions.

Publications