Progress 09/01/22 to 02/28/25

Outputs
Target Audience:Strawberry farmers, Cal Poly Strawberry Center Researchers, and Pest Control Advisors Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?This project provided one robot technician with training to operate a mobile robot. How have the results been disseminated to communities of interest?Results from this project were presented in tri-fold pamphlets to farmers and researchers. We also shared our results via email updated to farmers, researchers, and the USDA agricultural research service. 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. Automatic boom adjustment: Develop and test boom autonomy 1.1 Automatically adjust based on plant distance Description of research carried out: Downward and forward mounted sensors were added to measure canopy heights. Data collected: Plant distance relative to arrays were collected by multiple sensors; the boom control was adjusted in response to the new sensor feedback. Research findings and results: Two boom form factors were designed, built, and tested in open fields. Boom heights were compared and oscillations were reduced by controlling boom pistons and adding proportional valve controls. Zones were added over planted regions so the system could differentiate between plants and roads. 1.2 Automatically adjust based on geolocation Description of research carried out: Both geolocations and zones were tested. A new boom was developed to reduce lamp breakage. Data collected: Map data was collected along with the new format of treatment zones. Research findings and results: Combining automatic adjustment with zones in the field is an effective way to control for treatment regions in post-processing, improving the second pass through the field. This is more effective than automatically adjusting the boom in real-time because it resulted in less oscillations. 1.3 Automatically adjust based on obstacle detection Description of research carried out: Forward sensing was added to the boom position using intel real-sense cameras. We added obstacle identification and avoidance for human and obstacle scenarios. Research findings and results: Individual sprinklers and pipes in-row are difficult distinguish from tall weeds. Combining AI for feature location in the existing depth map can be an effective solution. 2. Dosing controller: Control dosing between multiple rows. 2.1 Establish dosing controller Description of research carried out: A wider boom has inherently less uncertainty, however it has less controllability. This results in lower variability overall and ultimately less oscillation. The multi-row controllability is a function of height and speed (no wings), speed is easier to control. Higher resolution encoders and proportional valve controls were added. Data collected: Dosing control data for both height and speed adjustment. Research findings and results: Speed adjustment can be controlled more reliably than then boom heights. Abrupt boom height adjustment led to bouncing. Boom adjustment must be done with a look-ahead distance proportional to vehicle speed. 2.2 Verify dosing uniformity between multiple rows Description of research carried out: To verify the dosing control over multiple rows, we used multiple high resolution UV sensors. Data collected: Multi-row dosing using multiple sensors simultaneously. Research findings and results: Although dosing variability was observed, implementation of the dosing controller led to safe dosing values across all rows. Initial wing designed led to burned plants. 2.3 Demonstrate commercial value Description of research carried out: We ran an experimental multi-row trial demonstrating effective dosing control at 650 J/m2, 1000 J/m2, and 1200 J/m2. Data collected: Data was collected in each section of the field to verify accurate dose. Research findings and results: A visible color difference was observed with healthier plants in the UVC section compared with chemical pesticides. Farmers have begun to adopt UV light on entire farms. 3. Experimentally validate treatment capabilities in openfields: Validation of the treatment 3.1 Evaluate autonomous treatment compliance Description of research carried out: Velocity and boom heights were studied before and after dosing control. Data collected: Height and velocity variability throughout the field. Research findings and results: We were able to successfully control the boom to a 12" plant spacing without causing damage over multiple rows. 3.2 Evaluate dosage estimates Description of research carried out: Dose was measured over multiple rows throughout the field. Data collected: Irradiance over time and dosing levels were recorded. Research findings and results: We achieved field dosing in ranges that control mildew, mites, and mold, in field conditions. 3.3 Collect and analyze data Description of research carried out: Fall and summer seasons were completed at Betteravia and Red Blossom farms. Data collected: Weekly data was collected from March until November in 2023 and 2024 across a total of 100+ acres of commercial strawberries. Research findings and results: We were able to demonstrate excellent results on mildew, mites, and botrytis (mold). We found that high dose was more effective on mites and botrytis, however, lower dose consistently controlled mildew. 3.4 Verify pest control in commercial field conditions Description of research carried out: Fall and summer seasons were completed at Betteravia farms. Data collected: Data on a control section was collected from March until November 2023 and 2024 Research findings and results: Data shows up to 90% kill-rate of mite eggs in a single pass and up to 46% Botrytis reduction. Mildew controls were completely replaced. 4. Establish safety protocols and pest control claims: Research safety and claims related to UV/autonomy. 4.1 Research requirements for establishing pest control claims Description of research carried out: UV-C safety and farm automation was researched. UV safety was consolidated on a tri-fold pamphlets and distributed to farmers and researchers at several agricultural events in 2023 and 2024. Research findings and results: Many farmers showed interest in UV technology; however, few were aware of the risks. The safety risk during treatment supported our effort for automation and we offered a word of warning to farmers who suggested building a system themselves. We offered farmers information on proper safety equipment and protocols. 4.2 Research and develop UV-C and automation safety protocols Description of research carried out: UV safety was distributed on a tri-fold pamphlets to farmers and researchers at several agricultural events in 2023 and 2024. Research findings and results: More farmers were aware of the safety risks of working with UV-C, however, there was still a need for education. Farmers saw the benefit of having automation to prevent the need for humans. Education was important to reduce fear of UV damaging plants. 4.3 Create training plan for robot operators and farmers Description of research carried out: New training materials were established and added to our online portal. 4.4 Incorporate safety improvements Description of research carried out: New automated power down and human tracking systems were added to ensure a safer work environment. Updates were made to the safety joystick for operators. Research findings and results: Automation added another level of safety for operators and allowed us to prescribe a minimum of 20 foot spacing between human operators and UV systems. 5. Complete cost estimates and final report In progress.

Publications

Progress 09/01/23 to 08/31/24

Outputs
Target Audience:Strawberry Farmers, Cal Poly Strawberry Center Researchers, and Pest Control Advisors Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?This project provided one full-time robot technician with training to operate a mobile robot. How have the results been disseminated to communities of interest?Results from this project were presented in tri-fold pamphlets to farmers and researchers. We also shared our results via email updated to farmers, researchers, and the USDA agricultural research service. What do you plan to do during the next reporting period to accomplish the goals?We plan to finalize reporting for our grant objectives.

Impacts
What was accomplished under these goals? 1.1 Automatically adjust based on plant distance Description of research carried out: Forward mounted sensors were added to measure canopy heights. Data collected: Plant distance relative to arrays were collected for the 2 rows in front of each camera; the boom control was adjusted in response to the new sensor feedback. Research findings and results: The new boom design performed much better than the previous winged boom. Proportional values made the boom easier to control. We tuned the new boom controller to smooth the boom vibrations. Zones were added over planted regions to eliminate sudden boom drops over roads. 1.2 Automatically adjust based on geolocation Description of research carried out: In place of geolocations, we created treatment regions that generalized better across large fields. The new boom was designed to lift completely away from plants, reducing bulb contact with the beds and less lamp breakage. Data collected: Map data was collected along with the new format of treatment zones. Research findings and results: Combining automatic adjustment with zones in the field is an effective way to control for treatment regions. Using satellite imagery, we can quickly construct these zones in the future, but interrow mapping remains a time-consuming challenge for large-tire in a narrow furrow. 1.3 Automatically adjust based on obstacle detection Description of research carried out: Forward sensing was added to the boom position using intel real-sense cameras. We added obstacle identification and avoidance for human and obstacle scenarios. Research findings and results: Individual sprinklers and pipes in-row are difficult to adjust for. In the future we plan to add AI-based identification for sprinklers, allowing for an overlaid depth map. Currently it is difficult to distinguish between tall weeds and sprinklers with depth alone. 2. Dosing controller: Control dosing between multiple rows. 2.1 Establish dosing controller Description of research carried out: The multi-row controllability was reduced when we moved to our new wide form-factor design, leaving only height and speed as controllable factors. New high resolution encoders and proportional valve controllers were installed and tested to smooth boom bouncing. Data collected: Dosing control data for both height and speed adjustment. Research findings and results: Speed adjustment can be controlled more reliably than then boom heights. Abrupt boom height adjustment led to bouncing. Boom adjustment must be done with a look-ahead distance proportional to vehicle speed. 2.2 Verify dosing uniformity between multiple rows Description of research carried out: To verify the dosing control over multiple rows, we used high resolution UV sensors that were developed by our team at TRIC. Data collected: Multi-row dosing using multiple sensors simultaneously. Research findings and results: Although dosing variability was observed, implementation of the new dosing controller maintained safe dosing values across all rows. 2.3 Demonstrate commercial value Description of research carried out: We ran an experimental multi-row trial demonstrating effective dosing control at 650 J/m2, 1000 J/m2, and 1200 J/m2. Data collected: Data was collected in each section of the field to verify accurate dose. Research findings and results: A visible color difference was observed with healthier plants in the UVC section compared with chemical pesticides. One of our trial farmers committed to UV treatment on their entire ranch in the following season. 3. Experimentally validate treatment capabilities in openfields: Validation of the treatment 3.1 Evaluate autonomous treatment compliance Description of research carried out: We compared velocity control before and after adding ultrasonic sensors for automatic boom control. Data collected: Height and velocity variability throughout the field. Research findings and results: We were able to prevent boom to plant contact over rough terrain with a 12" target spacing. 3.2 Evaluate dosage estimates Description of research carried out: Using a single remote sensor, we collected real-time dosing data to verify dosing in the field. This was done at night several times throughout the season to confirm dosing levels. Data collected: Irradiance over time and ultimately, dosing levels were recorded. Research findings and results: We achieved field dosing in ranges that control mildew, mites, and mold, effectively demonstrating the complete value described by USDA scientist in the lab. 3.3 Collect and analyze data Description of research carried out: Fall and summer seasons were completed at Betteravia. Data collected: Weekly data was collected from March until November across a total of 52 acres of commercial strawberries. Research findings and results: We were able to demonstrate excellent results on mildew, mites, and botrytis (mold). We found that high dose was more effective on mites and botrytis, lower dose consistently controlled mildew. 3.4 Verify pest control in commercial field conditions Description of research carried out: Fall and summer seasons were completed at Betteravia. Data collected: Data on a control section was collected from March to July Research findings and results: Data shows up to 90% kill-rate of mite eggs in a single pass. Botrytis was controlled more-so than many fungicide trails. Results were communicated regionally to farmers in Santa Maria and led to completely booking our fleet of robots. 4. Establish safety protocols and pest control claims: Research safety and claims related to UV/autonomy. 4.1 Research requirements for establishing pest control claims Description of research carried out: UV-C safety and farm automation was researched. UV safety was consolidated on a tri-fold pamphlets and distributed to farmers and researchers at several agricultural events in 2023. Research findings and results: We offered farmers information on proper safety equipment and protocols. Farmers saw automation as an effective risk reduction. 4.2 Research and develop UV-C and automation safety protocols Description of research carried out: UV safety was distributed on a tri-fold pamphlets to farmers and researchers at several agricultural events in 2024. Research findings and results: More farmers were aware of the safety risks of working with UV-C, however, there was still a need for education. Farmers saw the benefit of having automation to prevent the need for humans. This education induced a fear that UV is bad for plants, so we had to make it clear that appropriate dosing will not damage plants. 4.3 Create training plan for robot operators and farmers Description of research carried out: New training materials were established and added to our online portal for safety certification. 4.4 Incorporate safety improvements Description of research carried out: New automated power down and human tracking systems were added to ensure a safer work environment. Updates were made to the safety joystick for operators. Research findings and results: Automation added another level of safety for operators and allowed us to prescribe a minimum of 20 foot spacing between human operators and UV systems. 5. Complete cost estimates and final report Grant was extended to 2/28/2025. This objective will be completed at the end of the reporting period.?

Publications

Progress 09/01/22 to 08/31/23

Outputs
Target Audience:Strawberry Farmers, Cal PolyStrawberry Center Researchers, and Pest Control Advisors Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?This project provided one full-time robot technician with training to operate a mobile robot. How have the results been disseminated to communities of interest?Results from this project were presented in tri-fold pamphlets to farmers and researchers. We also shared our results via email updated to farmers, researchers, and the USDA agricultural research service. What do you plan to do during the next reporting period to accomplish the goals?There are many key activities that we will continue into 2024. We will complete new activities that we did not address in this reporting period (see outline above), but we will also expand on some of the previously completed activities. Importantly, we will test automation with a newly constructed boom and compare results with the wing design. We will add additional autonomy to improve dosing, safety, obstacle avoidance, and prevent bulb breakage. Using a set of sensors, we will simultaneously collect data over multiple rows to better compare variation between rows during treatment. Finally, we will complete an additional season, tracking improvements in dosing uniformity as a result of the multirow dosing controller.

Impacts
What was accomplished under these goals? 1. Automatic boom adjustment: Develop and test boom autonomy 1.1 Automatically adjust based on plant distance Activities completed: An array of ultrasonic range sensors were added to the boom for tracking plant distances below the UV system. Software was written to connect this information to GPS sensor data for relating distances to field geometries. We compared multiple boom configurations to determine if uncertainties could be reduced through physical design. Data collected: Plant distance relative to arrays were collected for each array; the boom control was adjusted in response to sensor feedback. Discussion: Testing for this objective led to a substantial redesign of the boom system. We found that the 3 row wings resulted in a flapping motion when the robot ran through the field. This natural vibration was a result of poor boom rigidity combined with the magnitude of the disturbances associated with a narrow track width. To settle the flapping motion, proportional control was added to the hydraulic valve controllers, then the boom and robot track width was redesigned to improve the rigidity and reduce disturbance. Calculations were completed to predict the impact of these adjustments. Changes in action: Boom was resigned and wheel spacing was increased to 5x to 320 inches, new data will be collected to compare this improved system with our previous boom design. 1.2 Automatically adjust based on geolocation Activities completed: Geolocations along the planned route were automatically distinguished based on treatment, travel, turn, and road states. These relationships were fed into the boom control. Data collected: Map data was collected along with automatically segmented transition zones. Discussion: Automatic boom adjustment must be tied into the geometry of the farm by adding geolocated tags for different regions. To distinguish between regions, maps were automatically segmented based on the action of the robot. Changes in action: For commercial viability, the robot must be equipped with enough heigh adjustment to exit steep rows, so the boom will be redesigned to lift above the wheels, well out of the way of the rows. 1.3 Automatically adjust based on obstacle detection Activities completed: No forward sensing was added to the boom during this reporting period, however, as mentioned in section 1.2 the boom does automatically adjust to changing terrain. Next, we plan to add forward sensors to avoid obstacles such as porta potties and fences. Discussion: It is worth noting the importance of autonomy. We damaged several lamps by manually operating the boom near the farm wind block fences. 2. Dosing controller: Control dosing between multiple rows. 2.1 Establish dosing controller Activities completed: By simulating variation in boom height vs terrain variability, we determined an acceptable range of treatment heights and boom movements to maintain minimum dosing variability. Data collected: Simulation data for boom responses using randomized plant heights. Discussion: We attempted dosing measurements in objective 2.2 to verify the effectiveness of our multirow dosing controller, however their resolution was too course to compare with simulation results. We plan to use multiple digital UV sensors in 2024. 2.2 Verify dosing uniformity between multiple rows Activities completed: To verify the dosing control over multiple rows, we used dosing cards spread across 7 rows as the robot passed over. Data collected: Low, medium, and high dosing measurements based on dosing cards. Discussion: It was difficult to compare dosing cards because of the coarseness of their resolution. It will be important to have higher resolution measurements to concretely verify dosing uniformity. Changes in action: Incorporate more accurate dosing measurements using multiple digital UV sensors. 2.3 Demonstrate commercial value N/A 3. Experimentally validate treatment capabilities in openfields: Validation of the treatment 3.1 Evaluate autonomous treatment compliance Activities completed: We compared velocity control before and after adding ultrasonic sensors for automatic boom control. Data collected: GPS positions which were plotted as velocities. Discussion: Although height and velocity data is relatively noisy, it gives us insights on how much irradiance (and therefore speed) we can add before the uncertainty results in either burned plants or ineffective dosing. We 3.2 Evaluate dosage estimates Activities completed: Using a single remote sensor, we collected real-time dosing data to verify dosing in the field. This was done at night several times throughout the season to confirm dosing levels. Data collected: Irradiance over time and ultimately, dosing levels were recorded. Discussion: We achieved dosing in ranges that were consistent with powdery mildew control. In 2024 we will target higher dosing to achieve mite and mildew levels. 3.3 Collect and analyze data Activities completed: Fall and summer seasons were completed at Red Blossom Data collected: Weekly data was collected from March until November across a total of 55 acres of commercial strawberries. Discussion: We were able to demonstrate excellent results on mildew. We started with a 25-acre plot in March, then moved the robot to a full block in July. The farmer did not need to see a control section on the second plot because we were able to demonstrate effective results on a large enough plot in the beginning of the year. 3.4 Verify pest control in commercial field conditions Activities completed: Fall and summer seasons were completed at Red Blossom Data collected: Data on a control section was collected from March to July Discussion: Data shows a substantial improvement in control between treatment and control sections. We started with mildew and now have strong confidence in effectiveness with this disease. UV was demonstrated as a complete replacement for chemicals for mildew control. NDVI data was also collected and along with in-field data, was presented to farmers, researchers at the Strawberry Center, and collaborators at the USDA Agricultural Research Center. 4. Establish safety protocols and pest control claims: Research safety and claims related to UV/autonomy. 4.1 Research requirements for establishing pest control claims N/A 4.2 Research and develop UV-C and automation safety protocols Activities completed: UV-C safety and farm automation was researched. UV safety was consolidated on a tri-fold pamphlets and distributed to farmers and researchers at several agricultural events in 2023. Discussion: Many farmers showed interest in UV technology; however, few were aware of the risks. The safety risk during treatment supported our effort for automation and we offered a word of warning to farmers who suggested building a system themselves. We offered farmers information on proper safety equipment and protocols. 4.3 Create training plan for robot operators and farmers Activities completed: Vishnu created a thorough safety training for robot operators and an accompanying online portal for documentation. All our robot operators were trained and signed off on their understanding of UV safety. Discussion: This documentation was critical because UV does not immediately cause discomfort. Safety equipment can be inconvenient. We used burnt strawberry plant leaves to show operators that UV has an indirect impact that is not immediately visible (similar to sunburn). 4.4 Incorporate safety improvements Activities completed: The UV system was analyzed for electrical safety. Components were replaced to improve electrical safety. We also focused on operator safety and tested a new safety rates joystick. Discussion: UV requires high power that can be dangerous if not isolated properly. Safety e-stops and breakers were added with lock-out protocols to prevent electrical risks. Additionally, we tested a new safety rated joystick. 5. Complete cost estimates and final report N/A

Publications