Desarrollo de un prototipo de un robot móvil terrestre autónomo para fumigación agrícola, en el cultivo de mora de castilla.

Abstract

Traditional agricultural fumigation applies a set of knowledge, techniques and tools that the farmer acquires through experience in crop management; however, fumigation with phytochemical products must be precise to have a uniform application throughout the crop and thus avoid resistance of pests. The curricular integration work presents the development of an autonomous prototype for agricultural fumigation, focused on the cultivation of blackberries from Castilla. The corresponding research established the characteristics of the prototype structure and the appropriate algorithms to meet the requirements focused on improving the traditional fumigation practice of blackberries from Castilla. To have a real-time perception of the prototype's environment, a LiDAR sensor is used. Its data is analyzed on the Raspberry pi by implementing ROS, Python and Raspbian, working with nodes that facilitate communication between software and hardware. In addition, to control the movement of the prototype, a multilayer perceptron neural network was implemented, which is responsible for predicting the speed and direction values of the "RAPFA" prototype according to the trajectory through the center of the path between rows that is generated in each iteration, so that it moves at a constant speed of 4km/h since at this speed drift is avoided because the drop receives more energy and better penetration into the crop occurs. The tests carried out reveal a saving of around 27% of the liquid applied to the crop in traditional fumigations with backpack pumps with a total range of 1.60m between rows. In this way, the farmer's work time is reduced by up to 73% according to the results obtained and prolonged exposure to chemical substances harmful to their health is minimized.