CROP SIGNALING FOR AUTOMATED WEED/CROP DIFFERENTIATION AND MECHANIZED WEED CONTROL IN VEGETABLE CROPS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: TERMINATED
• Funding Source: OTHER GRANTS
• Reporting Frequency: Annual
• Accession No.: 1004606
• Grant No.: 2014-51181-22379
• Project No.: CA-D-BAE-2270-OG
• Proposal No.: 2014-07909
• Program Code: SCRI
• Project Start Date: Sep 1, 2014
• Project End Date: Aug 31, 2019
• Grant Year: 2014

Project Director
Slaughter, D. C.

Recipient Organization
UNIVERSITY OF CALIFORNIA, DAVIS
410 MRAK HALL
DAVIS,CA 95616-8671

Performing Department
Biological and Agr. Engineer

Non Technical Summary
Stakeholders state that better weed management tools in vegetable crops are a high priority need and that methods to reduce or eliminate hand-weeding because of labor availability, competition, and cost are critically important for growers to remain profitable. Vegetable crops such as lettuce, onion and tomato are very susceptible to damage from weed competition. Weed management programs in vegetable crops generally cannot depend on herbicides alone. Herbicides used in vegetable crops help suppress weeds, however, an integrated multi-tactic weed management system is required that includes cultural practices such as crop rotation, as well as physical control tools like inter-row cultivation and hand weeding. Dependence upon hand weeding for weed control, a major cost component in the system, is an indication that the existing cultivator and herbicide technologies are not adequate for the task and that hand weeding remains an expensive but necessary component in vegetable weed management.There are few selective herbicides for vegetables and those available are 40 to 50 years old and control a limited number of weeds. Loss of old vegetable herbicides continues. Dust in farmworker housing in the Salinas Valley has been linked to increased exposure to pesticides used in agricultural applications. Research indicates that a structural shift in the agricultural labor pool is underway, and that fewer workers - including foreign-born workers - are likely to be available to work in low-skill low-wage jobs in the future. Rural education is on the rise, leading to higher levels of education. Higher levels of education are often associated with higher per capita incomes, and a shift away from low skill low wage work into higher wage jobs. Focused attention on alternative methods of weed control is especially important given the dwindling supply of agricultural workers in California and elsewhere in the US, competition for existing workers, and the concomitant upward pressure on wages. This project's long-term goals are:• To develop reliable weed control technologies that can accurately detect, locate and automatically kill weeds without damage to the crop using:o less labor,o reduced dependence on older high rate herbicides,o and at a substantially lower cost than current commercial practices. New technologies developed in this project will automate weed control tasks that are currently labor intensive (such as hand weeding) or that require herbicides. These new technologies have the potential to reduce:• hand weeding costs by 25% ($45 to $95 per acre depending upon the crop),• the negative environmental impact of chemical-based weed control, and• management problems associated with availability, competition, and cost of labor.The project objectives will also address the need to enhance efficiency of herbicide and labor inputs required, as well as to enhance the options for weed management systems in organic vegetable crops that have high weeding labor expenses.These solutions will help increase the competitiveness, profitability and long-term sustainability of the U.S. vegetable crop industry.

Animal Health Component

0%

Research Effort Categories

Basic

25%

Applied

75%

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
402	7299	2020	100%

Knowledge Area
402 - Engineering Systems and Equipment;

Subject Of Investigation
7299 - Research equipment and methods, general/other;

Field Of Science
2020 - Engineering;

Keywords

food safety

machine vision

mechanization

precision agri

sustainable weed control

vegetable crops

Goals / Objectives
The goal of this project is to use a trans-disciplinary systems approach for the development of automated, intra-row weed control solutions in vegetables. The project's long-term goals are to develop reliable weed control technologies for vegetables that can accurately and rapidly differentiate crop and weeds, determine their spatial location and automatically kill weeds without damage to the crop using:• less labor,• reduced dependence on older high rate herbicides, and at a substantially lower cost than current commercial practices.The project objectives also address the need to enhance efficiency of herbicide and labor inputs required, as well as to enhance the options for weed management systems in organic vegetable crops that have high weeding labor expenses.Three core technologies: 1) crop signaling, 2) machine vision, and 3) automated intra-row weed destruction, will be integrated using a systems approach, to create a novel, automated solution for weed control in specialty crops. The project is divided into seven objectives.1. Develop machine vision technology needed to map weeds and identify crop plants marked by signaling technology.2. Develop a robust vegetable crop signaling technology appropriate for use in vegetable fields.3. Develop a spray trajectory planner for dynamic control of the micro-spray herbicide applicator.4. Develop a high-speed and high resolution centimeter scale herbicide spray application system.5. Integrate the crop plant signaling and micro-spray technologies into vegetable production systems.6. Evaluate the efficacy and efficiency of the integrated system on intra-row weed control.7. Assess the economic, and obstacles to adoption of the technology.

Project Methods
This project uses a trans-disciplinary systems approach for intra-row weed control solutions in vegetables. Three core technologies: 1) crop signaling, 2) machine vision, and 3) automated intra-row weed removal, will be integrated using a systems approach to create a novel, automated solution for weed control in specialty crops. The project team will bring expertise in plant science, engineering, and economics to address the biological, technological, and economic challenges that need to be solved in order to create a robust solution that can succeed in bringing a sustainable weed control solution to vegetable crops growers.To achieve our solution for automated intra-row weed control the approach is divided into seven activities.1. Develop machine vision technology needed to map weeds and identify crop plants marked by signaling technology.2. Develop a robust vegetable crop signaling technology appropriate for use in vegetable fields.3. Develop a spray trajectory planner for dynamic control of the micro-spray herbicide applicator.4. Develop a high-speed and high resolution centimeter scale herbicide spray application system.5. Integrate the crop plant signaling and micro-spray technologies into vegetable production systems.6. Evaluate the efficacy and efficiency of the integrated system on intra-row weed control.7. Assess the economic, and obstacles to adoption of the technology.Field studies will be conducted to evaluate the strength and longevity of the marker material in lettuce, onion and tomato fields, and to evaluate the efficacy of weed control, weeding labor and impact on crop yield and quality. To assess the signal strength, field studies will be conducted in all three growing regions (AZ, CA, and WA) to evaluate the performance of the plant signaling technology under a wide range of vegetable crop growing conditions. Production situations will include drip, furrow, and sprinkle irrigation systems, solar exposure level due to geographic location and weather, and will study the stability and strength of the crop signals over the course of the growing season. Monitoring will be conducted with image processing technology. Data on signal strength will be subjected to repeated measures analysis to determine how long the signal lasts. Field experiments will be designed to determine if two passes of the intra-row precision sprayer improves weed control and labor use efficiency in lettuce and onion compared to current practices. The intra-row precision sprayer will be used for sequential applications on the same plots to accomplish the tasks of 1) thinning (lettuce only) and 2) intra-row weeding. The trial design will be a randomized complete block design with four replicates. Evaluations will include stand counts to evaluate the impact of the automatic weeder on the crop stand, visual vigor estimates, as well as time required for follow-up hand weeding and in lettuce for thinning.The primary economic aspects of the project will include capital and practice cost and profitability analyses of the proposed engineering solutions and weed control as compared to standard practices for the different crops. This will include a heightened focus on ownership costs, herbicide use, and labor inputs. Growers will be surveyed each year and their input used for economic and labor data collection. California will be the focus of the project's economic component in years one and two. During this time the research team will standardize the assessments and analyses and make appropriate adjustments, if necessary, to the engineering techniques and solutions. In years three and four the economic component will be expanded to include vegetable systems in the other states/areas of study. Secondary efforts will be directed at identifying adoption obstacles to adoption, if any, and potential solutions to overcome them. Budget Planner software will be used to generate the economic analyses, which will include equipment use, hand and machine labor, material input costs and interest on operating capital. Crop yield and quality (marketable versus cull product) will also be evaluated at the time of the analyses to determine profitability associated with each approach as compared to standard or traditional practices. Capital (ownership) costs for the new technology will be included in the analyses.A multi-faceted outreach effort is planned for this project that uses new approaches to technology transfer in ways that will accelerate awareness, acceptance and implementation of the new technologies and practices developed. It includes:• a powerful connection to national vegetable crop growers through our unique relationship with the American Vegetable Grower magazine,• utilizing the UC Davis Weed Research Information Center and its website (WRIC, http://wric.ucdavis.edu/ ) to announce events, post results and educational materials, and allow stakeholders to interact (ask questions, participate in our weed control blog, and enroll in educational programs); and• face-to-face communication through field visits, onsite demonstrations, and interactions at various industry related meetings.

Progress 09/01/14 to 08/31/19

Outputs
Target Audience:The target audiences reached by our outreach efforts in this reporting period primarily include members of the fresh market vegetable crop industry, both locally and trough our project wed page. Face-to-face communication efforts reached individuals in organizations such as the tomato grower members of the California tomato Research Institute, vegetable grower members of the California Leafy Greens Research Board and the Yuma Fresh Vegetable Association, representatives from vegetable seed producers, representatives from small equipment manufacturers, professional economists working in academic institutions, government agencies and departments, private industry and agribusiness, and non-governmental organizations, growers and members of the public who attended annual Weed Day events held at the University of California, Davis, and outcomes discussed with manufacturers, growers, researchers and general public in the Pacific Northwest. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?At UC Davis, two postdoctoral scholars, two graduate students and two undergraduate students were actively involved in this research project. The students received direct mentoring from the postdoctoral scholars and the faculty in the areas of engineering design and field research. The postdoctoral scholars and students were also engaged in professional development by attending seminars and conferences where they presented posters or oral presentations on their research contributions to the project's objectives. A postdoctoral scholar was involved in communicating with the Arizona State University on conducting integration experiments with the crop signaling sensing technology and the precision spray technology the UC Davis laboratory. At Arizona State University, two post-doctoral researchers worked on the project. They gained research experience and a better understanding of machine design and development. Two Engineering Technicians also worked on the project which enhanced their technical skills and knowledge of automated technologies for precision weed control. At Washington State University, one Ph.D. student, and three visiting scholars were actively involved in this project. One research scientist and one post visiting scientist coordinated the research tasks and helped with data analysis. PIs, scientists, students, and scholar interacted frequently to provide guidance to the student and scholar. Students and scholars carried out day-to-day research activities including concept development and discussion, prototype design and fabrication, and lab and field data collection. Students/scholar was also supervised for research paper writing, presentation, and publication. How have the results been disseminated to communities of interest?Arizona State University team applied research results and technical information was disseminated to over 1,000 (AZ numbers) growers, industry personnel, researchers and 750 members (AZ numbers) of the general public through presentations and demonstrations at field days (4), on-site demos (27), extension meetings (5) and ag festivals (1). Many more were reached through websites (1), YouTube video postings (3), and scientific (3), extension (1) and popular press (11) publications. Approximately 150 students were reached through formal classroom instruction (4) by hosting student groups (4). Washington State University team presented leveling system work at several conferences (e.g. 2018 Biorobotics Conference of International Federation of Automatic Control, 2018 Annual International Meeting of American Society of Agricultural and Biological Engineers), which attracted a lot of attention from peer researchers. One Ph.D. dissertation has been completed and two journal articles have been submitted for review. Fennimore delivered 5 outreach presentations during the final year of the project. This covered information regarding the strategies tested and results. The presentations were made to vegetable growers, consultants, and Farm Advisors. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? 1. Our team at the University of California, Davis has developed several machine vision technologies needed to map weeds and identify crop plants marked by crop signaling technology. One machine vision technique for robotic weed control based on plant label crop signaling technique was developed and experimentally evaluated in tomato and lettuce fields. Several trials were successfully conducted and the result shows 100% detection accuracy. A special in-field imaging chamber was equipped with ultra-violet lights and three sets of mirrors surrounding the crop plants to identify the stem soil-entry location of plants even when the plants leaned towards soil surface due to wind or rain in a field having densely populated weeds. One machine vision algorithm was developed to find the stem soil-entry location of tomato plants based on geometric appearance in the mirror images. For foliar crop signaling, an imaging chamber was developed with UV and White LEDs together to identify the weeds and crop plants for precision herbicide applications. The white light system recognizes all the plants including crops and weeds. The UV light illumination excites foliar crop signals to detect lettuce plants and distinguish them from weeds and provide a spray map to use with micro-jet spray systems. A fluorescence imaging system was developed to detect the uptake and translocation of systemic crop signaling in vegetable crops like lettuce, celery and snap bean. A machine vision algorithm was developed to detect celery crop plants and distinguished them from weeds based on the systemic crop signaling approach where plants were being treated with Rhodamine-B compound before transplanting. 2. A novel technique using crop signaling to detect crop plants and distinguish them from weeds for robotic weed control in leafy vegetables was successfully developed. Three different crop signaling systems were developed such as systemic markers, plant labels, and topical markers. In the systemic markers method, the seeds are coated with a signaling compound or a transplanted crop where the seedling plugs are treated at 1st true leaf stage prior to transplanting. Rhodamine-B was considered as marking compound and applied to vegetable crops such as lettuce, celery and snap bean. The marker was successfully observed after germination in the stems of bean seedlings at stages of the first leaf to multiple leaves over time in several vegetable crops. The plant labels are marked with fluorescence signaling compounds that are placed with the crop plants at transplanting in the field. The topical markers technique uses markers in which the signaling compounds are applied directly to the stem/foliage of crop plants before transplanting. In-field trials showed very promising results for plant labels and topical markers to classify weed/crop in the automatic weed control process. The methods have very low false-positive error rates and the classification accuracy achieved for plant labels and topical markers techniques are 100%, and 99.7%, respectively. 3. A dynamic model of a spray trajectory planner was developed and implemented, which computes the optimal positions of precision herbicide actuation events for a vegetable crop weeding robot. The planner takes consideration of the model-based mechanical design of the developed micro-jet herbicide sprayer. The algorithm takes into account the design of valve actuation as well as the spray fluid properties, weed proximity to the crop and travel speed to successfully target weeds with herbicide from a moving robotic weed control system. 4. A high speed, centimeter-scale resolution sprayer that can spot apply herbicides to weeds while traveling at speeds that are viable for commercial farming operations was developed at the University of Arizona. The performance of the device was evaluated in terms of spray delivery accuracy, off-target spray quantity, weed control efficacy, and crop safety. The spray assembly comprised 12 custom-built spray assemblies spaced 1 cm apart. The device was tested with lettuce in the laboratory at a travel speed on 2.0 mph while targeting four weed species at four stages of growth. Results showed that targeting accuracy of spray delivered was ± 2 mm and that the percentage of off-target spray was less than 3%. Weed control efficacy exceeded 95% and there was no observable crop injury. Information and knowledge gained was shared throughout the project via multiple outreach means described below. 5. Real-Time control of a herbicide micro-jet sprayer was successfully integrated with the weed-crop mapping system using crop signaling developed at UC Davis to create a complete autonomous weed control robot. The system uses a camera to acquire color images and the machine vision algorithm can distinguish between lettuce plants and weeds of many species and sizes based on crop signaling to create the weed-crop localization mapping. The pulse-jet micro-dosing actuator was controlled in real-time to target individual weeds by herbicides applications based on the mapping. Several experiments have been conducted with actual lettuce plants and different types of weeds. The experimental results show that the actuator can accurately at the target point at the speed of 3.2 km h-1. The Washington State team worked on integrating crop signaling, machine vision and precision spraying methods to develop a robotic system for weed removal in vegetable crops including onion and carrots. When commercially adopted, the weeding robot will dramatically alter the traditional ways of weed control in the vegetable fields. A research prototype was fabricated and evaluated in the lab and field conditions. Angular and position error data were collected for both laboratory and field conditions. The leveling system is able to maintain the angular error within 3 degrees in the field condition with a series of ditches (20 cm deep and 30 cm apart) when the forward speed was up to 51 cm/s. The leveling system is able to effectively reduce the angular error caused by complex and uneven field conditions. A machine vision system was developed to distinguish plants and weeds. A dual-camera-based end-effector position control system was fabricated and was integrated with a prototype weeding robot. A series of laboratory and field tests were conducted to estimate the accuracy of the end-effector to hit desired targets. End-effector position accuracy in reaching desired targets has been collected under laboratory and field conditions. The weeding robot achieved an end-effector position control accuracy of less than 2 mm when the robot was moving continuously on an uneven field surface at a speed of 0.45 ms-1. 6. The study evaluated the developed integrated system for intra-row weed control. The crop signaling system is very effective and efficient in detecting vegetable crop plants and distinguishing then from weeds to allow them to be removed successfully. The commercialization of the developed automatic weed control system is possible. 7. The goal of this research was to evaluate labor, yield and cost differences and impacts that are associated with standard (established) and automated (newer) weed management technologies for lettuce grown along the Central Coast of California. Practices have included standard mechanical inter-row implements (e.g, rolling and sled type cultivators), herbicide applications, and hand weeding in the past, and more recently the use of automated intra-row technologies or "intelligent cultivators". Research conducted on field stations and commercial farms from 2015 to 2018 indicates that automated cultivators can significantly decrease hand weeding labor, with no significant impact on yield. The report focuses on estimating differences in total weed management costs for lettuce, which includes ownership (investment) and operating (production) costs for hand weeding and automated cultivation techniques.

Publications

• Type: Journal Articles Status: Published Year Published: 2019 Citation: Raja R, Slaughter DC, Fennimore SA, Nguyen TT, Vuong VL, Sinha N, Tourte L, Smith RF, Siemens MC. Crop signalling: A novel crop recognition technique for robotic weed control. Biosystems Engineering. 2019 Nov 1;187:278-91.
• Type: Journal Articles Status: Published Year Published: 2019 Citation: Su WH, Fennimore SA, Slaughter DC. Fluorescence imaging for rapid monitoring of translocation behaviour of systemic markers in snap beans for automated crop/weed discrimination. Biosystems Engineering. 2019 Oct 1;186:156-67.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: Raja R, Slaughter DC, Fennimore S. A novel weed and crop recognition technique for robotic weed control in a lettuce field with high weed densities. In 2019 ASABE Annual International Meeting 2019 (p. 1). American Society of Agricultural and Biological Engineers.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: Rajaa R, Slaughtera DC, Fennimoreb S, Siemensc M. Precision weed control robot for vegetable fields with high crop and weed densities. In 2019 ASABE Annual International Meeting 2019 (p. 1). American Society of Agricultural and Biological Engineers.
• Type: Other Status: Published Year Published: 2019 Citation: Raja R, Slaughter DC, Fennimore S. Automatic crop recognition based on geometric appearance for a non-chemical robotic weed control system in a tomato field. In 2019 ASABE Annual International Meeting 2019 (p. 1). American Society of Agricultural and Biological Engineers.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: Su W. Automated Identification of Systemic Fluorescent Markers in Vegetable Seedling Leaves for Weed and Crop Differentiation. In2019 Annual Conference of the American Society for Horticultural Science (ASHS) 2019 Jul 23.
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: Su WH, Fennimore SA, Slaughter DC. Computer Vision Technology for Identification of Snap Bean Crops using Systemic Rhodamine B. In2019 ASABE Annual International Meeting 2019 (p. 1). American Society of Agricultural and Biological Engineers.
• Type: Journal Articles Status: Under Review Year Published: 2019 Citation: Raja R, Thuy T. Nguyen, Slaughter DC, Fennimore S. Real-time weed-crop classification and localisation technique for robotic weed control in lettuce. Biosystems Engineering. 2019 August
• Type: Journal Articles Status: Under Review Year Published: 2019 Citation: Raja R, Thuy T. Nguyen, Slaughter DC, Fennimore S. Real-time robotic weed knife control system for tomato and lettuce based on geometric appearance of plant labels. Biosystems Engineering. 2019 August
• Type: Journal Articles Status: Submitted Year Published: 2019 Citation: Raja R, Thuy T. Nguyen, Slaughter DC, Fennimore S. Real-time stem detection and classification from weeds for robotic weed control in processing tomato. Biosystems Engineering. 2019 August
• Type: Conference Papers and Presentations Status: Published Year Published: 2019 Citation: Fennimore SA, Cutulle M. 2019. Robotic weeders can improve weed control options for specialty crops. Pest Management Sci. 75:1767-1774. https://doi.org/10.1002/ps.5337
• Type: Theses/Dissertations Status: Submitted Year Published: 2019 Citation: Chen, L. 2019. A weeding robot auto-leveling system for typical vegitable fields in pacific northwest region. phd dissertation. washington state university.
• Type: Journal Articles Status: Accepted Year Published: 2019 Citation: SA. Fennimore, H. Kennedy, DC. Slaughter, T. Nguyen, V. Vuong, R. Raja. Detection of Marked Lettuce and Tomato by an Intelligent Cultivator and Effect on Weed Removal and Hand Weeding Requirements. Weed Technology, (2019)

Progress 09/01/17 to 08/31/18

Outputs
Target Audience:The target audiences reached by our outreach efforts in this reporting period primarily include members of the fresh market vegetable crop industry, both locally and trough our project wed page. Face-to-face communication efforts reached individuals in organizations such as the tomato grower members of the California tomato Research Institute, vegetable grower members of the California Leafy Greens Research Board and the Yuma Fresh Vegetable Association, representatives from vegetable seed producers, representatives from small equipment manufacturers, professional economists working in academic institutions, government agencies and departments, private industry and agribusiness, and non-governmental organizations, growers and members of the public who attended the 60th annual Weed Day event held at the University of California, Davis, and outcomes discussed with manufacturers, growers, researchers and general public in the Pacific Northwest. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?At UC Davis, two postdoctoral scholars, and one graduate student were actively involved in this research project. The students received direct mentoring from the postdoctoral scholar and the faculty in the areas of engineering design and field research. The postdoctoral scholar and students were also engaged in professional development by attending seminars and conferences where they presented posters or oral presentations on their research contributions to the project's objectives. A postdoctoral scholar was involved in communicating with the Arizona State University on conducting the indoor experiment with the herbicide sprayer at UC Davis laboratory. At Washington State University, one graduate student, one post-doctoral scholar, and one scientist were actively involved in this project. Research scientist coordinated the research tasks and helped with data analysis. PIs, scientist, postdoc, and student interacted frequently to provide guidance to the student and postdoc. Student and postdoc carried out day-to-day research activities including concept development and discussion, prototype design and fabrication, and lab and field data collection. Postdoc and students were also supervised for research paper writing, presentation, and publication. At the University of Arizona, Dr. Mark Siemens, and a Post-doc was engaged in research activities for the project in developing a high-speed and high-resolution centimeter scale herbicide sprayer and shipped to UC Davis to conduct the weed spraying experiment in real time. They have conducted 3-week teaching workshop on robotic and automated weeding systems at their institute in 2017 and in total 12 people have attended. Moreover, they have advised research manager on the state of current technologies for automated, precision weed control in Davis, California in 19th June 2018. How have the results been disseminated to communities of interest?We presented our research findings at the 2018Annual International Meeting of the American Society of Agricultural and Biological Engineers, which attracted a lot of attention from peer researchers. We also presented and demonstrated our work to various stakeholders (including researchers, manufacturers and growers) at annual meetings of the California Tomato Research Institute, the California Leafy Greens Research Board,the California Weed Society Conference, the Southwest Ag Summit, Yuma, Ariz., the Pre-Season Vegetable Workshop, Yuma, Ariz., Salinas Valley Weed School, Salinas, California, and the University of Arizona Engineering Camp - Sessions, Yuma, Ariz. During the reporting period, a combined audience of over 1,200 individuals were reached directly via the various outreach means listed below. Many more were reached through publications and popular press articles. What do you plan to do during the next reporting period to accomplish the goals?At UC Davis, we plan on continuing the development of a robust systemic crop signaling technology appropriate for use in direct-seeded vegetable crops (objective 2), continuing the design improvements in the system for direct application of the crop signaling material to the stems of upright crops like tomato, the integration of the machine vision technology developed for detecting and distinguishing crop plants from weeds with the high-speed weed destruction technology being developed at the University of Arizona,further evaluation of the efficacy and efficiency of the integrated system on intra-row weed control (objective 5, 6), and assessment of the economic and other obstacles to adoption of the crop signaling technology (objective 7). At the Arizona State University, we plan to develop a spray manifold comprised of 12 solenoid valves for precision in-row weeding. The device will span a width of 12 cm, with each solenoid valve/nozzle targeting a 1 cm wide area. The unit will be integrated with the imaging and control systems to form a precision weeding machine. System performance will be evaluated in a laboratory environment. We also plan to develop a spray assembly that can deliver two streams of spray per solenoid valve. We will further investigate optimal linkage assembly designs for the towed carriage assembly. At Washington State University, we plan to work on the integration of crop signaling and machine vision system (being developed by UC Davis team) on the current prototype of the robotic weeding system and to test the performance of the integrated system in vegetable fields in WA. Moreover, we plan to complete the economic analysis, which will includeinformation on investment and production costs and impacts on labor for both lettuce and tomatoes (CA and AZ). The current survey results do not include data from the Washington State vegetable industry. Additional data and analysis would further our understanding of the vegetable industry's use and needs, and the potential for adoption of automated technologies. The analysis will include a discussion about the economics, limitations of the analysis, and recommendations for the future.

Impacts
What was accomplished under these goals? Accomplishment under the goals are provided below 1. A novel technique using crop signaling to detect and classify weeds vs. crop plants for robotic weed control in leafy vegetables has been developed. The technique is simple and lowcost, in which the crop plants are marked with a unique machine-readable signaling compound before planting. Thereafter, the crop signal is used for rest of the season to automatically detect and classifyweeds and crop plants by the smart machine for automatic weed control. The algorithm was specifically developed for a vision based weed-spraying control system for lettuce. Although the technique can be adopted for many other annual specialty crops. In the machine vision technique, a vision sensor i.e., a portable camera, is mounted on the robot to capture the top-view image of the marked lettuce plants illuminated with ultraviolet (UV) lights. The crop signaling technique gives the crop plants the ability to produce unique optical signals when excited under the UV lights. This unique signal is used by the machine vision system to detect and precisely locate them. Another camera was mounted in front of the robot to capture the same top-view of the crop plants under the white light. Once both the UV and white images are captured for particular crop plants, the developed image processing algorithm process them to establish the plant map to precisely locate the position of the weeds and crop plants. An improvement in a previously developed algorithm to detect the main stem of tomato plants in the outdoor field has been made. The previous algorithm was confused whilefinding theposition of the rootof the tomato plantwhen the plant waslying down on the soil due to windy weather. The improved algorithm was successfully able to find the main stems of crop plants even when they were lying down on the soil. 2. Research has been progressing screening potential materials for the systemic marker crop signaling technology that is appropriate for use in vegetable seeds. Important molecular properties include lipophilicity, electrical charge and molecule size. In this experiment, many marking compounds have been evaluates to coat the seeds of crop plants and preliminary research indicated that Rhodamine B has been found as the most suitable one to date. Performance varies, however, with the type of vegetable seed. 3. A trajectory planning algorithm has been developed in Labviewfor dynamic control of the micro-spray herbicide applicator. The developed precision sprayassembly is spaced one centimeter apart delivers one stream of spray per solenoid valve. In total there are 12 solenoid valves covering a 12 cm total width in the lateral direction of the vegetable seedline. 4. Precision spray assemblies developed as part of this project deliver one stream of spray per solenoid valve. Each spray assembly is spaced one centimeter apart. To increase precision, investigations were made into the development of a spray assembly that would deliver two streams of spray per solenoid valve. This would increase the precision to 0.5 cm. Several designs were fabricated and tested including designs with different types of plunger seats (protruding, recessed), orifice diameters, orifice thru-hole lengths, and a distance between orifices. In total, six designs were evaluated. Unfortunately, none of the enhanced designs provided satisfactory results, primarily due to producing a too much off-target spray (spray on the crop is undesriable for an herbicide). We will conduct research on a potential solution during the next review period. Work was initiated on optimizing the design of the linkage assembly for the wheeled carriage (bed-follower) that will support the automated weeding machine's cameras and solenoid valve/nozzle assemblies. A linkage assembly capable of evaluating a parallel linkage and a tandem wheel linkage was mounted to a toolbar supported gauge wheels. Sensors were installed to measure the angular deflections of the carriage, toolbar, three-point hitch and tractor as the implement was pulled through the field. Preliminary results showed that carriage angle generally deviated less than 0.5 degrees, but deviations of 1 degree or more were common. The importance of this finding is that if the spray assembly were mounted 10 inches above the soil surface, a rotation of 1 degree would induce a targeting error of 0.44 cm. 5. Work has progressedtowards integrating the crop plant signaling and micro-spray technologies for weed control systemto increase vegetable production sustainability. In this system, the herbicide sprayer is placed 70 cm behind the camera in the weed control robotic system. Hence, thedevelopedalgorithm plans a trajectory for each solenoid valve over 12 x 70 cm area before the robot starts spraying weeds in the weed/cropmap. 6.The UC Davis team has conducted replicated field trials of the sensing and weed removal solutions for intra-row weed control for dicot vegetables (lettuce and tomato) at Salinas and UC Davis agriculture farm. Field research results confirm that crop signaling applied prior to planting greatly reduces the chance of false recognition of weeds as crops in the machine vision sensor. On-farm weed control results obtained in 2018 justify a minor change in focus with the prototype system toward mechanical weed control with results ranging in mechanized weed reduction from 50 to 95% and farm labor savings in the range of 50 to 75%. The weed/crop classificationaccuracy is measuredfor the algorithm developed for lettuce field is99.75% and, 98.11% of sprayable weeds detected with a detection time of 1.2 seconds per pair of images. This technique is highly accurate, reliable and robust than other sensor-based technique. 7. During 2017 the research team discussed the possible value of developing a questionnaire for vegetable growers and allied industry to determine the current use and potential for adoption of automated weed management technology. Research that examines the adoption of mechanized technology for field and tree crops exists in the literature; little or no comparable information has been found for vegetable crops, most notably vegetables grown for the fresh market. Historically, the development of technological solutions for fresh market vegetable production has been difficult because of the visual and tactile sensory attributes that humans bring to field practices and harvest decisions. To add to the literature, and as part of the Crop Signaling project, the research team developed a 10-question survey to discern current use and level of satisfaction for mechanized technologies in planting, thinning, weeding and harvesting vegetable crops. The survey also sought to determine the reasons why industry may or may not adapt existing or new vegetable technologies, and obstacles to their use. Questions related to general farm characteristics were also included to determine each respondent's primary role in the vegetable industry, farm size, crops grown and gross sales. The survey was granted exempt status by the UC Davis Internal Review Board on 9/17 and deemed exempt by the University of Arizona IRB, also on 9/17. The questionnaire was constructed, managed and data analyzed using Qualtrics Online Survey Software between August 2017 and August 2018. By design, the questionnaire was kept short to encourage greater industry participation, pilot tested, and first administered in-person at UC Cooperative Extension's (UCCE) Salinas Valley Weed School (SVWS) on 10/31/17. The survey was again administered in-person during the second meeting on 2/28/18, at UCCE's Central Valley Fresh Market Tomato Meeting (FMTM), and for a third time administered online between 3/14/18 and 4/1/18, targeting Arizona Vegetable Growers.

Publications

• Type: Theses/Dissertations Status: Submitted Year Published: 2018 Citation: Kennedy, H. "Weed Removal Efficacy and Labor Implications of an Intelligent Cultivator in Vegetable Crops Marked with Machine Vision Detectable Signals", University of California, Davis, 2018.
• Type: Conference Papers and Presentations Status: Published Year Published: 2018 Citation: Kennedy H, Fennimore SA, Rachuy JS, Slaughter D, & Nguyen TT, "Weed vs. Crop Differentiation Using Crop Marking Systems", California Weed Science Society, January, 2018.
• Type: Conference Papers and Presentations Status: Published Year Published: 2018 Citation: Kennedy H, Fennimore SA, Rachuy JS, Slaughter D, & Nguyen TT, "Weed Control Efficacy and Labor Implications of an Intelligent Cultivator in Vegetable Crops Marked with Machine Vision Detectable Signals", Weed Day University of California, July, 2018.
• Type: Conference Papers and Presentations Status: Published Year Published: 2018 Citation: Kennedy H, Fennimore SA, Rachuy JS, Slaughter D, & Nguyen TT, "Weed Removal Efficacy and Labor Implications of an Intelligent Cultivator with Crop Marking Systems", American Society of Agricultural and Biological Engineers, July, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Tourte, L. "The economic components of the crop signaling for vegetables", vegetable grower group, November, 2017.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Tourte, L. "The economic components of the crop signaling for vegetables", vegetable grower group, February, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Tourte, L., "The economic components of the crop signaling for vegetables", The joint Western Agricultural Economics Association / Western Extension Farm Management Committee meeting, Anchorage Alaska, June 25, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Fennimore SA and Slaughter D. "Weed management in lettuce", California Leafy Greens Research Program, Salinas, CA October 10, 2017.
• Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Fennimore SA, "Intra-row weed control automation in California vegetable crops", Salinas Valley Weed School, University of California Cooperative Extension, Salinas. CA October 31, 2017.
• Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Fennimore SA, "Physical weed control", West Hills College, Coalinga, CA, November 3, 2017.
• Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Fennimore SA, "Intra-row weed control automation in California vegetable crops", University of California Cooperative Extension, Santa Maria, CA, November 28, 2017.
• Type: Conference Papers and Presentations Status: Other Year Published: 2017 Citation: Fennimore SA, "Weed Control Automation, the best path forward for improved weed management in Vegetable Crops", Association of Applied IPM Ecologists, Visalia, CA, November 30, 2017.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Automation, detection and pest management", BASF Corporation, Monterey, CA, February 22, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Weed management in lettuce", California Leafy Greens Research Program. Pismo Beach, CA March 27, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Weed removal automation", Weed Workgroup, University of California Agriculture and Natural Resources, Ontario, CA April 9, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Weed removal automation", Agroecology Workgroup, University of California Agriculture and Natural Resources, Ontario, CA April 12, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Weed removal automation", Innovation Summit, University of California Agriculture and Natural Resources, Davis, CA April 23, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Ag Labor, hand weeding, Ag Chem & Weed Control Automation", Weed Science Seminar, UC Davis Plant Science Dept, Davis, CA. May 7, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Crop signaling technology for weed control in vegetable crops", International Congress of Crop Protection Chemistry, Ghent, Belguim, May 22, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Fennimore SA, "Automation, detection and pest management", BASF Corporation, Marina, CA, June 26, 2018.
• Type: Conference Papers and Presentations Status: Other Year Published: 2018 Citation: Kennedy HJ, Fennimore SA and Slaughter D, "Weed versus crop differentiation using a crop marking system with an intelligent cultivator", Weed Day, UC Davis Plant Science Dept, Davis, CA, July 12, 2018.
• Type: Conference Papers and Presentations Status: Published Year Published: 2018 Citation: Chen L, Karkee M, He L, Wei Y, and Zhang Q, "Evaluation of a Leveling System for a Weeding Robot under Field Condition", 6th International Federation of Automatic Control Conference on Biorobotics, Beijing, China, July 12-15, 2018.
• Type: Conference Papers and Presentations Status: Awaiting Publication Year Published: 2018 Citation: Chen L, Karkee M, He L, Wei Y, and Zhang Q, "Design and Laboratory Evaluation of a Leveling System for a Conceptual Robotic Weeder", Annual International Meeting of American Society of Agricultural and Biological Engineers, Detroit, Michigan, USA, July 31-Aug 2, 2018.
• Type: Other Status: Published Year Published: 2018 Citation: Loren, Christina, "Precision Spray Assembly", Market Day Report - Live Interview, Nashville, TN: Rural Media Group Inc. RFD-TV, August 10, 2018.
• Type: Other Status: Published Year Published: 2018 Citation: Karst T, "Research looks at automating weed killing function", Lenexa, Kansas: Farm Journal, Inc., August 7, 2018. URL: https://www.farmjournalagtech.com/article/research-looks-automating-weed-killing-function.
• Type: Other Status: Published Year Published: 2018 Citation: Downing, J, "Next-generation mechanization", Calif. Agr. 72(2):102-104, 2018.
• Type: Other Status: Published Year Published: 2017 Citation: Varelas T, "Tecnolog�a agr�cola a�n no reemplaza la mano humana en los campos" Chronkite Noticas Arizona PBS 30 November, 2017. Phoenix, Ariz.: Cronkite News. URL: https://www.youtube.com/watch?v=iWxTsU77xlg.
• Type: Conference Papers and Presentations Status: Awaiting Publication Year Published: 2018 Citation: Siemens, M.C, "Agricultural Engineering", University of Arizona  USDA STEAM Residential Summer Program, Yuma, Ariz., August 2, 2018.

Progress 09/01/16 to 08/31/17

Outputs
Target Audience:The target audiences reached by our outreach efforts in this reporting period primarily include members of the fresh market vegetable crop industry, both locally and trough our project web page. Face-to-face communication efforts reached individuals in organizations such as the tomato grower members of the California Tomato Research Institute, vegetable grower members of the California Leafy Greens Research Board and the Yuma Fresh Vegetable Association, representatives from vegetable seed producers, representatives from small equipment manufacturers, growers and members of the public who attended the 60th annual Weed Day event held at the University of California, Davis., and outcomes discussed with manufacturers, growers, researchers and general public in the Pacific Northwest. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?At UC Davis, one postdoctoral scholar, one graduate student and two undergraduate students were actively involved in this research project. The students received direct mentoring from the postdoctoral scholar and the faculty in the areas of engineering design and field research. The postdoctoral scholar and students were also engaged in processional development by attending seminars and processional conferences where they presented posters or oral presentations on their research contributions to the project's objectives. At Washington State University, one PhD student and one visiting scholar were actively involved in this project. One research scientist coordinated the research tasks and helped with data analysis. PIs, scientist, student and scholar interacted frequently to provide guidance to the student and scholar. Student and scholar carried out day-to-day research activities including concept development and discussion, prototype design and fabrication, and lab and field data collection. Students were also supervised for research paper writing, presentation and publication. At the University of Arizona, Dr. Mazin Saber, a Post-doc, was engaged in research activites ifor the project n the Fall 2016 - Winter 2016 period and benefitted from direct mentoring by faculty in Arizona. How have the results been disseminated to communities of interest?We presented our research findings at the 2017 Annual International Meeting of American Society of Agricultural and Biological Engineers, which attracted a lot of attention from peer researchers. We also presented and demonstrated our work to various stakeholders (including researchers, manufacturers and growers) at annual meetings of the California Tomato Research Institute, the California Leafy Greens Research Board, the 2017 CPAAS technology expo (July 31th, 2017), the California Weed Society Conference, the Southwest Ag Summit, Yuma, Ariz., the Pre-Season Vegetable Workshop, Yuma, Ariz., Salinas Valley Weed School, Salinas, Calif., and the University of Arizona Engineering Camp - Sessions I & II, Yuma, Ariz. During the reporting period, a combined audience of over 1,250 individuals were reached directly via the various outreach means listed below. Many more were reached through publications and popular press articles. What do you plan to do during the next reporting period to accomplish the goals?At UC Davis in year 4, we plan on continuing the development of a robust systemic crop signaling technology appropriate for use in direct seeded vegetable crops (objective 2), continuing the design improvements in the system for direct application of the crop signaling material to the stems of upright crops like tomato, the integration of the machine vision technology developed for detecting and distinguishing crop plants from weeds with the high-speed weed destruction technology being developed at the University of Arizona, further evaluation of the efficacy and efficiency of the integrated system on intra-row weed control (objective 6), and assessment of the economic and other obstacles to adoption of the crop signaling technology (objective 7). At the University of Arizona in year 4, we plan to develop a spray manifold comprised of 16 solenoid valves for precision in-row weeding. The device will span a width of 16 cm, with each solenoid valve/nozzle targeting a 1 cm wide area. System performance will be evaluated in the lab. The unit will also be integrated with the imaging and control systems to form a precision weeding machine. The precision weeding machine will be evaluated in field tests where its performance in terms of weed control efficacy, crop injury and yield will be assessed. We also plan to conduct research to improve the performance of a spray system for marking lettuce seedlings with fluorescent paint during thinning. Finally, we plan to continue work on optimizing the design of the linkage assembly for the wheeled carriage (bed-follower) that will support the automated weeding machine's machine vision cameras and solenoid valve/nozzle assemblies. Three designs, a parallel linkage, a tandem wheel linkage and a sled design will be investigated. The units will be instrumented to measure the angular deflections of the links as the carriages move through the field. The data will be analyzed using kinematic analysis tools in SolidWorks™ to design a linkage assembly with minimal lateral and rotational movement under normal operating conditions. At Washington State University, in year 4 we plan to work on the integration of crop signaling and machine vision system (being developed by UC Davis team) on current prototype of the robotic weeding system, and to test the performance of the integrated system in vegetable fields in WA.

Impacts
What was accomplished under these goals? Stakeholders state that better weed management tools in vegetable crops are a high priority need and that methods to reduce or eliminate hand-weeding because of labor availability, competition, and cost are critically important for growers to remain profitable. This project uses a trans-disciplinary systems approach for intra-row weed control solutions in vegetables. Three core technologies: 1) crop signaling, 2) machine vision, and 3) automated intra-row weed removal. The UC Davis team has created machine vision technology needed to map weeds and identify crop plants marked by signaling technology (objectives 1 and 3). Research results from in-field trials conducted in 2016 and 2017 show that the novel UC Davis machine vision technology developed in this project is very successful in detecting and distinguishing crop plants from weeds using the foliar spray implementation of the crop signaling system. Results show that both false positive crop recognition errors and false negative errors are very low (well below 2% and 1%, respectively), even under very dense weed populations sometimes found in organic farms with high weed pressures. These results are quite good, and strongly validate the research hypothesis that the crop signaling concept can create a machine-readable crop plant with a unique optical signature that greatly simplifies and ensures the success of the crop vs. weed detection task. UC Davis also created a more robust crop signaling technology for automated weed detection in vegetable crops (objective 2). The work accomplished addresses the project objective for a novel application method for a short-term crop signaling compound that is required to complete the systems approach to the design of the automatic weed control system. Knowledge obtained from replicated in-field trials on the UC Davis campus farm show that: 1) increasing the target size of the crop signaling material applied to the crop increases the robustness in crop detection during the critical period of weed competition (3 weeks post planting), 2) decreasing the leakage of material to the soil increases the robustness in crop detection due to a significant reduction in false positive signal from the soil, and 3) that the crop signaling material identified is non-toxic to vegetable crop plants, like tomato, when applied at an effective dose at planting. New knowledge regarding the photostability of the crop signaling material was obtained that will help researchers understand and further improve the design of the automated machine. An improved design of the high-speed weed destruction technology for mechanical (i.e. non-chemical weed control, objective 4) was also created at UC Davis. This new design incorporates features for automatic depth control of both the sensing system and the weed removal mechanism as well as features for better integration of the intra-row and inter-row weed removal mechanisms. For the objective 5, a custom version of the UC Davis weed control system was created for the unique aspects of the leafy vegetable context. The system was tested in replicated in-field trials in 2016 and 2017 and the results demonstrate that design improvements addressed the needs of adaptation of of sensing and weeding technologies for this local context. The UC Davis team has conducted replicated field trials of the sensing and weed removal solutions for intra-row weed control for dicot vegetables (lettuce and tomato, objective 6). Field research results confirm that crop signaling applied prior to planting greatly reduces the chance of false recognition of weeds as crops in the machine vision sensor. On-farm weed control results obtained in 2017 justify a minor change in focus with the prototype system toward mechanical weed control with results ranging in mechanized weed reduction from 50 to 95% and farm labor savings in the range of 50 to 75%. For objective 7, cost and return analyses conducted provide details on representative weed management programs for each specialty vegetable crop studied and also show relevant recent economic trends in the farm gate values and share of harvested acreage. Notable increases were seen, especially in value of production, for the majority of the vegetable crops. Weed control consistently ranks as one of the top ten challenges to the economic sustainability that organic vegetable crops producers face. Research at UC Davis shows that in evaluating total weed management costs by crop, hand weeding represents 36 % of the total in tomatoes (the lowest) to 92 % in organic spinach (the highest). Washington State University team is working on objective 5, for weed removal in onion and carrots to develop a robotic platform to follow straight and curved paths with a maximum position error of 2.2 cm. The prototype platform is capable of following a pre-defined trajectory within desirable accuracy. Weed identification rate was lower (~30%) in the first two weeks after germination of carrots, which improved to ~80% after 14 days of germination. The weed detecting accuracy increased substantially for carrots after 14 days as carrot leaves started to form their unique shape around that time. In the onion field, the initial weed identification rate was lower than 30% in the first two weeks, which increased to 60% after 17 days of growth. Research at the University of Arizona has focused on developing a second spray assembly for precision weeding (objective 4). The unit was designed to deliver continuous spray to centimeter wide targets of various lengths at commercially viable travel speeds with minimal drift. Results showed that targeting accuracy for all treatments was better than ± 2.0 mm and ± 1.6 mm in the longitudinal and lateral directions respectively. Average percent coverage of the target areas was 59.3% and off-target drift was 4.9%. These results exceeded the established success criteria of delivering spray at the centimeter level scale of accuracy with < 5% off target spray. The spray assembly developed in FY17 differs from the FY18 design in that it utilizes a diverging-converging orifice and is designed to deliver individual droplets rather than a continuous spray stream. For all trials, average targeting accuracy was less than ± 2.1 mm and ± 1.7 mm in the longitudinal and lateral directions respectively. Average percent coverage of the target area (1 cm2) was 16.6% and off-target drift was 1.8%. These results exceeded the established success criteria of delivering a < 1x1 cm spray pattern with < 5% off target spray. Based on these results it was concluded that the two spray assemblies were best suited for different purposes. The thru-hole orifice design had better coverage, but more off-target spray than the diverging-converging design. It is therefore best suited for use with contact herbicides. The converging-diverging design with well-defined droplets and minimal off-target drift is better suited for use with systemic herbicides. The University of Arizona team also worked on identify a solenoid valve/nozzle combination that would deliver fluorescent marking paint to 3/4" diameter lettuce seedlings with good coverage and minimal drift (objective 2). Over 25 combinations were evaluated at different pressures, travel speeds and operating heights with several combinations that produced reasonable patterns were identified. To facilitate testing, a prototype automated thinner was modified to simultaneously thin lettuce seedlings and deliver marking paint to the "saved" crop plant. The performance of the solenoid valve/nozzle assemblies was reasonable with some combinations producing roughly 2x2" patterns, with minimal off target spray. Results showed that improvements in targeting accuracy and spray coverage are needed in order for the imaging system to reliably detect marked crop plants.

Publications

• Type: Journal Articles Status: Published Year Published: 2016 Citation: Fennimore, S.F., Slaughter, D.C., Siemens, M.C., Leon, R.G. & Saber, M.N. (2016). Technology for Automation of Weed Control in Specialty Crops. Weed Tech. 30(4), 823-837.
• Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: Chen, L., S. L. He, M. Karkee, and Q. Zhang. 2017. Weed Identification Rate Analysis for Different Growing Period. 1017 ASABE International Annual Meeting, July 17-19, Spokane, WA.
• Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: Nguyen, T.T., D.C. Slaughter, S.A.Fennimore, V.L. Vuong. 2017. Designing and evaluating the use of crop signaling markers for fully automated and robust weed control technology. ASABE Annual International Meeting 1700160.(doi:10.13031/aim.201700160) July 17-19, Spokane, WA.
• Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: V.L. Vuong, D.C. Slaughter, T.T. Nguyen, S. A. Fennimore, D. K. Giles. 2017. An Automated System for Crop Signaling and Robotic Weed Control in Processing Tomatoes. ASABE Annual International Meeting 1700160.(doi:10.13031/aim.201700160) July 17-19, Spokane, WA.
• Type: Other Status: Published Year Published: 2017 Citation: Schwartz, F.A. 2017. University of Arizona integrates farming and technology. AZ Business Magazine. April 7. Phoenix, Ariz.: AZ Big Media. Retrieved from http://azbigmedia.com/ab/university-arizona-farm-tech.
• Type: Other Status: Published Year Published: 2017 Citation: Dahl, H. 2017. UA partners to develop weed-killing bot. The Daily Wildcat 110(80): 8.
• Type: Other Status: Published Year Published: 2017 Citation: Schwartz, F.A. 2017. High-tech camera gets into the weeds in farm country. UA News. Tucson, Ariz.: Office of University Relations, Communication, University of Arizona. Retrieved from https://uanews.arizona.edu/story/hightech-camera-gets-weeds-farm-country.
• Type: Other Status: Published Year Published: 2016 Citation: Johnson, B. 2016. Researchers work to improve mechanized weeding. Ag Alert, November 30, p. 7, 9. Sacramento, Calif.: California Farm Bureau Federation
• Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: Slaughter, D.C. S.A. Fennimore, J.S. Rachuy, H. Pheasant. 2017. Weed vs. Crop Differentiation using Crop Marking Systems. California Weed Society Conference. Monterey, California January 18, 19, & 20, 2017.

Progress 09/01/15 to 08/31/16

Outputs
Target Audience:The target audiences reached by our outreach efforts in this reporting period primarily include members of the fresh market vegetable crop industry, both locally and trough our project web page. Face-to-face communication efforts reached individuals in organizations such as the tomato grower members of the California Tomato Research Institute, vegetable grower members of the California Leafy Greens Research Board and the Yuma Fresh Vegetable Association, representatives from vegetable seed producers, representatives from small equipment manufacturers, growers and members of the public who attended the 59th annual Weed Day event held at the University of California, Davis., and outcomes discussed with manufacturers, growers, researchers and general public in the Pacific Northwest. Changes/Problems:In the past year, we put efforts on a self-leveling system development. For the vegetable crop fields in WA, the crop plants are planted in high density with weeds growing very close to crop plants, which requires the weeding end-effector operating very precisely. However, closer investigation of field condition indicated that the path which the weeding robot drives through is uneven with a maximum level difference of 20 cm and a maximum distance of 38 cm between peaks. The rough driving surface will induce noticeable variation in roll and pitch angles of the robot (up to ± 10º on both pitch and roll directions) while driving. Therefore, it was essential to develop and evaluate a self-leveling system so that a stable and precise base can be provided for the weeding end-effector. What opportunities for training and professional development has the project provided?The project has provided state-of-the-art training for engineering students at the undergraduate and graduate levels and post-doctoral scholars in the design and development of advanced machine systems for on-farm mechanization in US agriculture. Students carried out day-to-day research activities including concept development and discussion, prototype design and fabrication tasks. Students were also supervised for research paper writing, presentation and publications. How have the results been disseminated to communities of interest?In the 2016 Precision Farming Expo held in Kennewick, WA (Jan 7-8, 2016), we demonstrated the robotic platform self-leveling system and presented a poster, which attracted a lot of attention from peer researchers, manufacturers and growers. We also communicated with local growers and conducted experiments in their fields. These growers were highly interested in our work, and are keen to see weeding robots working in the field. We are scheduled to present our work at AgriControl 2016, the 5th IFAC Conference on Sensing, Control and Automation for Agriculture (August, 14-17, 2016, Seattle, WA), which is expected to initiate more discussion and dissemination of information to various stakeholders. Project scientists shared results with automated thinning machine manufacturer Vision Robotics Inc. (San Diego, CA). Based on findings, advised them on improved solenoid valve/nozzle assembly designs for precision spraying of herbicidal materials for thinning lettuce. Project scientists shared results and demonstrated equipment in one-on-one meetings with numerous growers and industry representatives including Tanimura & Antle, Top Flavor Farms, JV Farms, Harrison Farms, Sudduth Farms, T&P Farms and Mellon Farms. What do you plan to do during the next reporting period to accomplish the goals?In year 3, field trials will be conducted in California to evaluate the efficacy and efficiency of the crop signaling based automatic weed control systems developed to date. We plan to conduct formalized tests of the solenoid valve/nozzle body(s) developed. Research effort will be initiated to develop a spray trajectory planner for dynamic control of the micro-spray herbicide applicator based upon the machine vision results obtained by the imaging system developed at UC Davis. We also plan to initiate work to optimize the design of the linkage assembly for the wheeled carriage (bed-follower) that will support the automated weeding machine's machine vision cameras and solenoid valve/nozzle assemblies. Two designs, a parallel linkage and a tandem wheel linkage, will be investigated. The units will be instrumented to measure the angular deflections of the links as the carriages move through the field. The data will be analyzed using kinematic analysis tools in SolidWorks™ to design a linkage assembly with minimal lateral and rotational movement under normal operating conditions. We also plan to initiate integration of the spray system, existing bed-follower and imaging software. We will also install the leveling system on the current field scale platform, test and evaluate the performance of the leveling system in the field condition. A robotic weeding prototype will be developed to work in Washington crops by integrating the power system, crop signaling and machine vision systems (being developed by UC Davis team), and a weeding end-effector (being developed by University of Arizona team) with the current field platform. A manuscript will be written and submitted to a peer-reviewed journal.

Impacts
What was accomplished under these goals? Stakeholders state that better weed management tools in vegetable crops are a high priority need and that methods to reduce or eliminate hand-weeding because of labor availability, competition, and cost are critically important for growers to remain profitable. Vegetable crops such as lettuce, onion and tomato are very susceptible to damage from weed competition. Weed management programs in vegetable crops generally cannot depend on herbicides alone. Herbicides used in vegetable crops help suppress weeds, however, an integrated multi-tactic weed management system is required that includes cultural practices such as crop rotation, as well as physical control tools like inter-row cultivation and hand weeding. Dependence upon hand weeding for weed control, a major cost component in the system, is an indication that the existing cultivator and herbicide technologies are not adequate for the task and that hand weeding remains an expensive but necessary component in vegetable weed management. There are few selective herbicides for vegetables and those available are 40 to 50 years old and control a limited number of weeds. Loss of old vegetable herbicides continues due to a variety of reasons. For example, dust in farmworker housing in the Salinas Valley in California has been linked to increased exposure to pesticides used in agricultural applications causing renewed concerns about the long-term sustainability of chemical pesticide use in agricultural production systems. In addition, research indicates that a structural shift in the agricultural labor pool is underway, and that fewer workers - including foreign-born workers - are likely to be available to work in low-skill low-wage jobs in the future. Rural education is on the rise, leading to higher levels of education. Higher levels of education are often associated with higher per capita incomes, and a shift away from low skill low wage work into higher wage jobs. Focused attention on alternative methods of weed control is especially important given the dwindling supply of agricultural workers in California and elsewhere in the US, competition for existing workers, and the concomitant upward pressure on wages. The project is divided into seven objectives. The accomplishments achieved in the second year of the project on each of these objectives are as follows: 1. An improved version of the machine vision system created to map weeds and identify crop plants in leafy green vegetables (e.g. lettuce) was developed in the past year. The new approach increased the number of viewpoints and improved upon the choice of vantage points to provide a more robust and effective system for identification of leafy green vegetables marked by crop signaling technology. The machine vision weed mapping approach was also revised to provide more consistent image quality and superior performance in weed detection and mapping. On-farm experiments were conducted in Salinas CA to evaluate the system performance in leafy green vegetable fields. The machine vision system for mapping weeds and identify crop plants in tomato fields was also improved and we determined that many of the same multiview features successfully used in the system created for leafy vegetable crops were of value in tomato as well. Field experiments in both tomato and leafy vegetable crops demonstrated that the new approach was extremely effective in crop plant identification and mapping when the crop plant labeling approach to crop signaling was used. 2. An automatic system for topical application of the crop signaling compound was developed. This mobile system automatically applies the crop signaling compound to the crop plants at planting. In addition, significant progress has been achieved in the development of the systemic crop signaling technique. Laboratory experiements show a very high success rate for establishing the crop signaling compound in leafy green vegetable crop plants when using the new method. 3. Effort on this goal will be conducted pending outcomes from objectives 1, 2, and 4 that will define the engineering design constraints and needs of the trajectory planner for micro-spray control. 4. The capabilities of the linear rail system fabricated in year 1 for evaluating solenoid valve/spray nozzle assemblies were enhanced. The micro-controller for controlling carriage speed and nozzle activation was upgraded for improved performance. The micro-controller was programmed to monitor carriage travel distance and to deliver droplets to targets at specified locations. This is an improvement over the previous system since it is the type of control system planned for use with the integrated intra-row weeding machine. In year 2 we developed our own design for the high resolution centimeter scale herbicide spray application system patterned after micro-dispensing valves used for devices such as ink-jet printers. When used in combination with an economically priced solenoid valve (< $30), system performance was excellent. The assembly was evaluated in a test where droplets were delivered to 9 targets spaced over a distance of 20 cm. Nozzle tip to target distance was 5" and travel speed was 2 mph. The average targeting accuracy was ± 0.35 mm and ± 1.3 mm in the longitudinal and lateral directions respectively. Average percent coverage of the target area (1 cm2) was 47.6% and off target drift was 1.8%. These results exceeded the established success criteria of delivering a < 1x1 cm spray pattern with < 5% off target spray. 5. Our focus is on integrating crop signaling, machine vision and precision spraying methods to develop a robotic system for weed removal in vegetable crops including onion and carrots. A self-leveling system was designed for robotic platform so that weeding end-effector remains level even when the driving path is uneven in the field, as is common in onion and carrot fields in the Pacific Northwest. A laboratory scale leveling platform encompassing multiple sensors and actuators was fabricated. The pitch and roll error of the leveling plate is less than +/- 0.25 degrees when the input pitch and roll angle varied from -8 degrees to 8 degrees. This level of accuracy in leveling will help keep the end-effector positioning error caused by uneven ground surface to a minimal level. 6. Field trials were conducted in Davis, CA and Salinas CA to evaluate the intra-row weed control performance of the automated weed control machine designed and fabricated at UC Davis. The Davis trials were conducted in tomato fields, and the Salinas trials were conducted in lettuce fields. Prelimnary results from these trials establish the technical feasibility of the crop signaling concept. Crop plant recognition was nearly 100% and the weed control efficacy was very good. 7. An economic analysis of the value of different types of vegetable crops grown in California was conducted. This economic analysis study was used to help guide the research priorities of the project in terms of identifying the most economically important vegetable crops to study with respect to the development of the systemic crop signaling compound. In 2014, the combined production value of head, leaf, and romaine lettuce was $2.0 billion, the highest value of any vegetable crop. Tomatoes (fresh and processing) followed at $1.6 billion, with broccoli ($807 million), carrots ($574 million), and peppers ($498 million) rounding out the top five places. These results directed the research effort in systemic crop signaling to focus on the leafy vegetable crops in year 2.

Publications

• Type: Websites Status: Published Year Published: 2016 Citation: http://ucanr.edu/sites/roboticweedcontrol/
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Vuong, V.L., T.T. Nguyen, S.A. Fennimore, D.C. Slaughter. 2016. Automated application of a crop signaling compound for automated weeding. Presented at the ASABE Annual International Meeting, Orlando, FL. July 17-20, 2016.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Siemens, M.C. & Norton, E.R. 2016. Automated Technologies for In-row Weeding, 2016 Southwest Ag Summit, Yuma, Ariz., February 25, 2016.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Chen, L., S. Kaewkorn, L. He, Q. Zhang, and M. Karkee. 2016. Design and evaluation of a levelling system for a weeding robot. Proceedings of the 5th Conference on Sensing, Control and Automation for Agriculture, August 14-17, Seattle, WA.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Chen, L., S. Kaewkorn, M. Karkee, L. He, and Q. Zhang. 2016. A prototype of leveling system for weeding robot. Precision Farming Expo, January 7-8, Richland, WA.

Progress 09/01/14 to 08/31/15

Outputs
Target Audience:The target audiences reached by our outreach efforts in this reporting period primarily include members of the fresh market vegetable crop industry, both locally and trough our project web page. Face-to-face communication efforts reached individuals in organizations such as the tomato grower members of the California Tomato Research Institute, vegetable grower members of the California Leafy Greens Research Board and the Yuma Fresh Vegetable Association, representatives from vegetable seed producers, representatives from small equipment manufacturers, growers and members of the public who attended the 59th annual Weed Day event held at the University of California, Davis., and outcomes discussed with manufacturers, growers, researchers and general public in the Pacific Northwest. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?The project has provided state-of-the-art training for engineering students at the undergraduate and graduate levels in the design and development of advanced machine systems for on-farm mechanization in US agriculture. Students carried out day-to-day research activities including concept development and discussion, prototype design and fabrication tasks. How have the results been disseminated to communities of interest?The the results been disseminated to communities of interest in this reporting period primarily by face-to-face communication with individuals in organizations such as the tomato grower members of the California Tomato Research Institute, vegetable grower members of the California Leafy Greens Research Board and the Yuma Fresh Vegetable Association, representatives from vegetable seed producers, representatives from small equipment manufacturers, growers and members of the public who attended the 59th annual Weed Day event held at the University of California, Davis., and outcomes discussed with manufacturers, growers, researchers and general public in the Pacific Northwest. Additional dissemination of results has been accomplished trough our project web page. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Stakeholders state that better weed management tools in vegetable crops are a high priority need and that methods to reduce or eliminate hand-weeding because of labor availability, competition, and cost are critically important for growers to remain profitable. Vegetable crops such as lettuce, onion and tomato are very susceptible to damage from weed competition. Dependence upon hand weeding for weed control, a major cost component in the system, is an indication that the existing cultivator and herbicide technologies are not adequate for the task and that hand weeding remains an expensive but necessary component in vegetable weed management. There are few selective herbicides for vegetables and those available are 40 to 50 years old and control a limited number of weeds. Loss of old vegetable herbicides continues due to a variety of reasons. In addition, research indicates that a structural shift in the agricultural labor pool is underway, and that fewer workers - including foreign-born workers - are likely to be available to work in low-skill low-wage jobs in the future. Focused attention on alternative methods of weed control is especially important given the dwindling supply of agricultural workers in the US, competition for existing workers, and the concomitant upward pressure on wages. This project's long-term goals are: To develop reliable weed control technologies that can accurately detect, locate and automatically kill weeds without damage to the crop using: o less labor, o reduced dependence on older high rate herbicides, o and at a substantially lower cost than current commercial practices. New technologies developed in this project will automate weed control tasks that are currently labor intensive (such as hand weeding) or that require herbicides. These new technologies have the potential to reduce: o hand weeding costs by 25% ($45 to $95 per acre depending upon the crop), o the negative environmental impact of chemical-based weed control, and o management problems associated with availability, competition, and cost of labor. The project objectives will also address the need to enhance efficiency of herbicide and labor inputs required, as well as to enhance the options for weed management systems in organic vegetable crops that have high weeding labor expenses. These solutions will help increase the competitiveness, profitability and long-term sustainability of the U.S. vegetable crop industry. The project is divided into seven objectives. The accomplishments achieved in the first year of the project on each of these objectives are as follows: 1. Develop machine vision technology needed to map weeds and identify crop plants marked by signaling technology. Three types of machine vision systems for identifying crop plants (vs. weeds) and to map weeds were developed. Machine vision system #1 was developed as a portable, lightweight, and completely self-contained system. The system was used outdoors in actual lettuce farms in California to characterize several engineering design parameters for crop identification and weed mapping tasks, and to collect in-field research data on the suitability, efficacy, and robustness of various crop signaling technologies being developed as part of objective 2). This system was successfully used during the 2015 growing season in Salinas CA and research data was collected on actual lettuce plants subjected to a variety of crop signaling technologies, grown outdoors on a farm. This data facilitated research and assessment tasks for objective 2. Machine vision system #2 was developed for identifying lettuce plants (vs. weeds) and to map weeds in commercial lettuce fields. The system was self-contained and accommodated commercial double row lettuce beds and used high-resolution color cameras mounted to detect both the target crop signaling technologies being developed as part of objective 2, and to view and map all the weeds growing in the 4 square foot zone surrounding the double row of lettuce seedlings centered on the bed. In-field experiments were conducted in Salinas CA during the 2015 growing season to evaluate the efficacy of the crop signaling technologies developed in objective 2 when applied under actual commercial production conditions and subjected to the natural outdoor growing environment of a commercial farm. Machine vision system #3 was developed for identifying tomato plants (vs. weeds) and to map weeds in commercial tomato fields. This system incorporated a novel machine vision approach that was customized for tomato plants and designed to successfully resolve the unique machine vision challenges that this crop's architecture presents to the crop signaling concept. 2. Develop a robust vegetable crop signaling technology appropriate for use in vegetable fields. A joint effort of collaborative study has been conducted between the project's partners, AgInnovations and the UC Davis Seed biotechnology Center, and the project investigators into identifying compounds that are suited for use as crop signaling technologies for the purpose of crop plant identification. Several compounds were evaluated and three materials were identified that met the performance objectives of: sufficient signal to noise ratio to allow highly accurate discrimination between the crop signal and weeds, is non-toxic when applied to crop plants, and has a minimum signal effectiveness lifetime of six to eight weeks. Different formulations of the materials were also evaluated and their environmental robustness (e.g. resistance to signal degradation) was compared in the outdoor farm environment for the target lifetime of eight weeks post germination. Two implementations of the crop signaling technology were successfully developed and tested in lettuce fields in Salinas CA, and three implementations of the technology were successfully developed and tested in tomato fields in Davis CA in the 2015 growing season. In-field, signal degradation rates for each implementation method were characterized and new knowledge regarding the optimal formulations of each material type was determined. 3. Develop a spray trajectory planner for dynamic control of the micro-spray herbicide applicator. Effort on this goal was delayed pending outcomes from objectives 1, 2, and 4 that will define the engineering design constraints and needs of the trajectory planner for micro-spray control. 4. Develop a high-speed and high resolution centimeter scale herbicide spray application system. A motorized linear rail system was fabricated for evaluating solenoid valve/spray nozzle assemblies. The unit is equipped with a micro-controller for controlling carriage speed and nozzle activation. Several commercial and custom solenoid valve/spray nozzle assemblies were evaluated using water sensitive paper. The best performing assemblies were able to produce oval shaped spray patterns having diameters less than 1 cm. Initial assessment of off-target spray (drift), however, exceeded design specifications, requiring additional engineering effort in this objective for mitigation. 5. Integrate the crop plant signaling and micro-spray technologies into vegetable production systems. A robotic weed control vehicle was designed based on the requirements established from the field measurements of cropping system including bed widths, intra-row crop plant spacing and variation in field condition including terrain. A self-leveling system has also been designed to maintain the orientation of weed-removal end-effector irrespective of the roll and pitch angles of the base platform. 6. Evaluate the efficacy and efficiency of the integrated system on intra-row weed control. Effort on this goal was delayed pending outcomes from objectives 1, 2, 4, and 5. 7. Assess the economic, and non-technical obstacles to adoption of the technology. Effort on this goal was delayed pending outcomes from objectives 1, 2, 4, and 5.

Publications

• Type: Websites Status: Published Year Published: 2015 Citation: http://ucanr.edu/sites/roboticweedcontrol/
• Type: Other Status: Published Year Published: 2015 Citation: Andrade, K.G. 2015. Planting Seed of the Future: Yuma Ag Center Making Inroads with Agricultural Technology. Yuma Daily Sun, August 30, p. 1, 3. Yuma, Ariz.: The Sun.