Files
Abstract
Weed control in agricultural fields poses a major challenge for farmers due to the significant losses that can be caused by weeds, impacting productivity. Herbicides have been the most effective weed control solution, in addition to other conventional methods like mechanical and manual weeding. However, the emergence of herbicide-resistant weed populations, concerns regarding herbicides' negative impact on the environment, and the labor-intensive nature of these conventional weed control methods threaten the ability of the agricultural industry to keep up with the growing demand. Furthermore, evidence shows that labor costs in agriculture are increasing rapidly due to labor shortages. All these factors necessitate alternative weed control solutions that are less labor-reliant, cost-effective, with minimal impact on the environment and crop health. Lasers have emerged as a viable weed control solution due to their capacity to precisely target weeds. Small autonomous robotic platforms employing lasers can be used to target weeds in their early growth stages and eliminate them.In this study, an autonomous robot that employed a diode laser for weed elimination was designed and implemented. The robot utilized a combination of visual servoing for motion control, the Robot Operating System (ROS) for coordination, and a Finite State Machine to manage its states, actions, and transitions. Data from sensors such as GPS, encoders, stereo cameras, and IMUs were utilized for weed detection, robot navigation, and control of the robotic manipulator arm. An Extended Kalman Filter (EKF) sensor fusion algorithm was employed to fuse sensor data for robot localization. Furthermore, the robot employed deep learning for weed detection, weed tracking, and path detection between cotton rows. It then used the Dynamic Window Approach (DWA) path planning algorithm for navigation. A 2D Cartesian manipulator arm was used to position the laser diode attached to a rotating pan-and-tilt mechanism for precise weed targeting. Experiments conducted in a cotton field showed that the robot was able to effectively navigate autonomously between cotton rows, detect and track weeds, and eliminate them with laser beam treatment. These results provide strong evidence of the feasibility of autonomous weed elimination using low-cost diode lasers on small robotic platforms.