OPERATION ALGORITHM FOR THE AUTONOMOUS ROBOTIC WEED MONITORING SYSTEM

Abstract

An autonomous robotic complex designed for monitoring weed vegetation requires an efficient algorithm for proper movement in the work area. (Research purpose) The research purpose is creating and testing an algorithm using a virtual robot model in Gazebo simulator and field tests using a prototype. (Materials and methods) It was noted that the developed algorithm includes a method of classification and decision-making. Indicated that at the first stage the robot is being prepared, including checking the systems and sensors. Then parallel programs are launched, including receiving and processing the video stream from the camera, detecting weeds and correcting the path traveled. (Results and discussion) It was determined that the camera continuously sends images of plants under the robot for processing. The second program detects weeds, and according to the data obtained, the robot’s path is corrected and the working part is positioned. It was established that in case of detection of weed vegetation, its coordinates are determined, which is necessary for localization in the work area and correction of the path. Then the path of the robot’s movement between the rows of plants is determined, and the error of deviation from the set course is transferred to the next step of the algorithm. (Conclusions) As a result of the experiments conducted in simulation and field tests, the algorithm showed the effectiveness and correct operation of the robot in real time. The created algorithm does not require significant computational resources, and its successful testing in various conditions has confirmed its applicability for monitoring weeds.