Design, development, and verification of a robotic spraying system and study on the application strategies for coverage optimization during site-specific chemical application

Abstract

Typically, farmers use large sprayers to conduct a uniform rate whole field application for pest management without spatial pest severity knowledge. This approach leads to untimely application, inadequate coverage, and off-target spray. Large self-propelled sprayers also have inherent application rate accuracy concerns, challenges to manage boom height, and little to manage wind speeds in order to minimize drift potential. Hence, new novel concepts are needed for site-specific pesticide applications targeting critically infested plants to reduce the chemical input, decrease the negative environmental impacts, and sustainably maximize the food yield. Robotic systems have shown tremendous potential to solve complex issues in the agricultural domain; however, systems lack which utilize a full stack of sensor and automated technologies to provide liquid application solutions to mitigate existing concerns. So, the overall intent of our research was to develop a robotic spraying system capable of site-specific chemical application to operate within crop rows and canopy, validate the sub-systems for accuracy and response time, test the performance of the overall spraying system for in-field application requirements, and quantify spray coverage using actual crop canopy structure in the greenhouse. To fulfill this objective, a four-wheel-drive differential steering mobile platform capable of maneuvering within 30 inches row crop spacing was developed to support necessary system components. In addition, a PWM-based individual nozzle-controlled spraying system having two solid vertical booms and six nozzles to conduct chemical applications was designed, built, and integrated with the robotic platform. The control system for both the platform and the sprayer was developed and validated through laboratory testing. Finally, an autonomous platform with a chemical application system was operated in a simulated crop environment in the greenhouse to study the spray coverage utilizing different application rates, emitter orientations, and platform operating strategies. The results demonstrated that the sprayer could maintain system pressure within ±5% of the target irrespective of the duty cycle and the number of active nozzles. The solenoids operated as intended, and the nozzle application pressure remains within ±5% of the target application pressure when operating one, three, and all six nozzles at 40% duty cycle. Also, the nozzle application pressure in two different booms and at three boom heights indicate no significant difference suggesting the developed system can spray uniformly and with a highly accurate application rate. The spray coverage data showed the highest overall deposition of 17.33% at both 15 GPA application rate and regular pass in 0° nozzle orientation (0BCT) and 15 GPA application rate and regular pass in 45° nozzle orientation (4BCT). The average spray coverage obtained was 26%, 16%, and 11% under 0BCT, and 21%, 19%, and 13% under 4BCT in the top, medium, and bottom canopy heights. In both 0BCT and 4BCT configurations, the spray coverage was higher (21.5%) in the inner lateral canopy region near the plant stalk than in the middle of the leaves (13%). The spray penetration data showed significant deposition of 21-56% of the total on plant canopy not directly facing the autonomous vehicle travel path, suggesting the capability of the sprayer to apply chemicals with sufficient droplet travel through crop canopies to provide coverage on the side of the crop not directly facing the sprayer. This study has provided valuable insights into the extent of spray drop coverage, which is significantly more than what is usually realized using over the canopy sprayers, and provided observations for potential system design improvements for great chemical deposition in the canopy in future. Future research should include computer vision to autonomously detect the pest incidence and severity to provide decision feedback to the sprayer for site-specific chemical application. Autonomous path-planning of the rover for navigating it through the row-crops is another crucial work for seamless chemical application. Further, I recommend field testing of the robot in other crops by incorporating the findings of this research.