Source: TRANSCEND ENGINEERING AND TECHNOLOGY, LLC submitted to
IRRIGATION MANAGEMENT USING NOVEL LOW-COST HIGH-RESOLUTION SENSING FOR INCREASED ADOPTION
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
1030006
Grant No.
2023-33530-39241
Cumulative Award Amt.
$181,395.00
Proposal No.
2023-00770
Multistate No.
(N/A)
Project Start Date
Jul 1, 2023
Project End Date
Feb 29, 2024
Grant Year
2023
Program Code
[8.4]- Air, Water and Soils
Project Director
Farrington, S.
Recipient Organization
TRANSCEND ENGINEERING AND TECHNOLOGY, LLC
1453 VT ROUTE 107
BETHEL,VT 05032
Performing Department
(N/A)
Non Technical Summary
Through our work with researchers and growers we have found that cost is the most significant barrier to adopting high-resolution soil water monitoring needed to meaningfully reduce agricultural water consumption. We previously developed, patented, and commercialized an innovation using Time Domain Reflectometry that quantifies in situ soil water content in highly resolved spatial profile, providing greatly improved understanding of the distribution and fate of applied water as a function of the irrigation schedule, but cost has been a barrier to adoption. The recent massive proliferation of wireless accessories has driven the cost of integrated circuit-based programmable RF transceivers unprecedentedly low, and their usability unprecedentedly high.Phase I of this proposed project will determine if it is feasible to base the design and operation of a spatially dense, multi-level soil moisture sensor around the exploitation of commodity RF electronic components to infer soil water content from soil dielectric properties. The work also seeks to quantify the tradeoffs of accuracy and cost that the approach presents. Phase-II work would build the algorithmic basis for a decision support system, the framework for which is under commercial development.A hyper-advanced decision support system is under development by our commercialization partners. Incorporating our resulting sensing innovation will span the "adoption gap," from data, to insights, to data driven recommendations and decisions. The result will be unprecedented efficiency in the management and conservation of water used in irrigated agriculture, with concomitant benefits for soil health and groundwater quality leading to greater sustainability.
Animal Health Component
37%
Research Effort Categories
Basic
35%
Applied
37%
Developmental
28%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
1117410202050%
1120210205050%
Goals / Objectives
The goal of the Phase 1 work is to establish the feasibility of the technical approach outlined above for lowering the cost of high-fidelity soil moisture monitoring by fivefold. Phase I technical objectives towards this goal include:Demonstrate a relationship between software retrievable RF transmission quality metrics and Ka of the environment.Quantify the accuracy, precision, and repeatability of the relationship in 1 and investigate the role of EC in this relationship.Confirm the expected cost savings of the envisioned device described.The Phase I effort will answer the following specific technical questions:Does VWC of the surrounding medium exert a systematic effect on the RF performance characteristic between pairs of low-cost IC-based RF transceivers?What are the specific relationships of software-retrievable metrics of signal transmission between RF chips to dielectric permittivity and EC of the surrounding medium?Are relationships in 2 sufficiently precise and repeatable to match the performance of existing commercial multi-level soil water monitoring products?Can a device based on these principles achieve a fivefold reduction in cost relative to existing commercial alternatives?Phase II objectives will include: developing, fielding, and validating a production prototype of the resulting high-fidelity VWC sensor, and demonstrating a deployment-to-decision workflow and technology framework that produces actionable irrigation management recommendations.
Project Methods
The project will involve an 8-month Work Plan comprising four tasks. Methods with regard to each task are described below.1.1.Task 1 - Develop Electronics for ExperimentationConducting experiments to answer the three technical questions posed will require an integrated electronic system to test, and the ability to control the environment enveloping the system. This task, to develop the device used to evaluate RF chip response to changes in the environment, will begin with chip selection. A wide variety of IC-based RF transceivers are available from a plethora of manufacturers, each designed to implement one or more specific industry protocols for wireless. Many include powerful microprocessors in the same silicon. Under this task we will review the devices for the signal metrics required of the protocols they implement, as well as manufacturer documentation for the most ubiquitous and affordable chips in each class to determine any non-standard signal indicators the manufacturer exposes to developers. Consideration will be given to RF frequencies vis-à-vis the penetration potential into soil. We will also consider the availability of eval boards or dev kits to speed development of the test system, and will select three to five of the hundreds of available transceiver/protocol/frequency (TPF) implementations to support initial evaluation.Next, we will develop a hardware/software prototype capable of addressing a variety of chips to perform the needed testing. We will connect and program a single board computer (SBC) as a controller to coordinate transceiver pairing and to retrieve and record signal data during the experiments. The physical configuration and form factor will be compatible with the test apparatus described below. Transceivers pairs will be separated by 5 cm to match the spacing envisioned for the final product.1.2.Task 2 - Experimentation and AnalysisBench testing and analysis are the central activities of the Phase I project. Task 2 will begin with design and construction of the apparatus that will be used to surround pairs of transceivers in materials of known, controlled dielectric and conductivity.The apparatus will comprise concentrically arranged tubes and/or cylinders, the inner of which will accommodate insertion and withdrawal of PCBs incorporating different pairs of transceivers and will provide isolation from the surrounding fluids. The outer "environmental vessel" will contain the fluid standards surrounding the transceiver tube. The inner diameter of the transceiver tube will closely mimic the form factor of the VWC probe we ultimately envision. The design will utilize commonly available construction and packaging materials. Once the apparatus is designed, we will procure materials and supplies and construct it in our Bethel, VT facility.Fluid standards will be used to control the dielectric environment in which signal readings are taken. Some materials that can be readily procured and provide a useful range of relative permittivity with negligible EC include corn oil (4), ethanol, (25), methanol (33), and de-ionized (DI) water (80). Later, well-sorted quartz sand will be saturated with different pore fluids, and the bulk Ka and EC of the mixture will be independently determined using conventional TDR which is more reliable than theoretical mixing models. This will also allow us to characterize the sensitivity of IC signal metrics to bulk EC by manipulating the conductivity of pore water a priori using KCl commercial conductivity standards diluted by DI.We anticipate evaluating the signals of each of the candidate TPFs in at least five pure liquid and five liquid-in-soil controlled permittivity materials (having negligible but measured EC), and in at least five water-in-soil materials having constant bulk permittivity and variable EC each. This combination will enable thorough evaluation of relationships between multiple indicators of signal character exposed by the chip manufacturers and variations in Ka (VWC) and EC. Precise control of the orientation and placement of transceivers within the tube using a 3D printed fixture will enable a high degree of repeatability.The microprocessor will be able to sample RSSI, LQI and other host controller parameters dozens or hundreds of times per second and compute average readings representing any reasonable time scale of relevance to operational VWC monitoring. Collected data will be organized in tabular files for subsequent data analysis.Data Analysis will begin with visualization of the acquired data from each candidate TPF using colored heatmaps of how each software-retrievable signal metric varies on a 2-D field of medium EC versus Ka, as a quick and intuitive way to look for systematic response of signal metrics to variation in the electrical properties of surrounding material. We will also produce scatterplots and compute Pearson correlation coefficients between pairs of variables.Different protocols may assign the same signal metric name (e.g., RSSI) to different definitions, as may different manufacturers. In addition, antenna designs and internal signal pre-processing methods will vary. We may therefore see different relationships in the data from different TPFs even within the same protocol. However, we expect LQI to be computed independent from RSSI and to relate to packet transmission metrics that have more to do with framing issues, edge preservation, etc. than to signal amplitude (i.e., RSSI). RSSI should clearly have a relationship to signal amplitude, and LQI should have a relationship to phase changes and/or the quality of edges in the modulation of the signal. In the simplified view of the problem, we will have two measures to solve for two unknowns. But there may be interplay. This is why candidate chip selection in Task 1 will emphasize the availability of additional metrics of signal character. Techniques we apply might include: multiple linear regression, multivariate multiple linear regression, canonical correlation analysis, multivariate adaptive regression splines, and partial least squares regression and random forests if warranted by the data complexity and dimensionality. The accuracy of models of material dielectric and EC as a function of multi-variate TPF signal metric response to the environment will be quantified, compared, and judged using coefficient of determination (R2) and root mean square error (RMSE) as measures of fit. Prediction accuracy of VWC determined from Ka and EC will be compared to the performance of leading commercial multi-level WVC probes that we have previously characterized. A model resulting from any TPF that matches or exceeds the accuracy of existing commercial probes will constitute success in achieving Objective 2.1.3.Task 3 - Production Cost EvaluationUnder this task, we will develop a complete design for a manufacturable probe and evaluate the production cost at volume to validate our expectations about the cost savings achievable by using commodity wireless communications technology. State of the art design tools will be employed, and a complete bill of materials (BOM) will be developed for each subassembly. The BOMs along with CAD models and design drawings will form a basic for obtaining estimates of materials and production costs from vendors and contract manufacturers.The cost to produce for the probe will be built up to account for additional industry normative costs of distribution, inventory, sales, marketing, and support. Final projected retail cost will be compared to the benchmark of $500 we have established, as well as to the cost of known "competitors" to evaluate the viability of the go-to-market strategy. A halving of any cost metric relative to the mean of commercial alternatives presented in our proposal will constitute successful attainment of Objective 3

Progress 07/01/23 to 02/29/24

Outputs
Target Audience:The target audience included irrigation companies, certified irrigation consultants, and certified agronomic consultants with whom we engaged through contact by anintermediary TABA consultant who arranged video teleconferences with roughly a dozen individuals deeply involved in the agricultural irrigation industry. In addition we reached researchers and policymakers through attendance and presentation at the National Soil Moisture Workshop. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The product provided opportunities for several staff and contractors to gain new experience interfacing software and hardware with commodity RF transceiver chips;planning, designing and conducting experiments;acquiring and analyzingdata, and drawing conclusions based on their research findings. How have the results been disseminated to communities of interest?The target audience included irrigation companies, certified irrigation consultants, and certified agronomic consultants with whom we engaged through contact by anintermediary TABA consultant who arranged video teleconferences with roughly a dozen individuals deeply involved in the agricultural irrigation industry. In addition findings have been accepted for presentation at the National Soil Moisture Workshop to be held in Logan, Utah in July 2024. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? The work began with simultaneous development of a sensing apparatus and an environmental apparatus. The sensing apparatus securely and reliably held required electronic components in geometric position relative to each other, an enclosing facsimile probe body, and surrounding media, enabling the repeatable collection of RSSI and SNR data from multiple time-varying and static environments representing a range of dielectric permittivity that is spread evenly over the range associated with dry to saturated soils. While this work failed to establish a consiustent enough relationship of RSSI and SNR from BLE and Zigbee chips to use tham as soil moisture sensors, similar adjunct work under a separate project found that CSI data from WiFi transceivers can effectively enable commodity priced WiFi chips to be used as soil moisture sensors. A simple random forest machine learning (ML) model was found to identify which media dielectric standard the transceivers produced CSI from with remarkable accuracy and consistency. A simple artificial neural network (ANN) ML model was found to perform estimation of the natural logarithm of apparent dielectric permittivity, ln(Ka), with mean square error (MSE) around 0.135, corresponding to an error in estimation of volumetric water content (VWC) of about 2%, competitive with the accuracy of existing commercial probes. A cost to produce analysis predicated on a detailed preliminary design and bill of materials (BOM) estimated cost to produce of about $431 dollars for a nominal 1-m multi-level sensor (MLS) probe using the new technology, enabling market entry at $800/unit, highly competitive with existing MLS soil moisture probesAll three objectives of the Phase I project were achieved.

Publications