A Mobile Robot with Deep Learning for Monitoring Coffee Farms

Resumen

Coffee is one of the most important products for Honduras, not only due to the amount of exports, but also its importance in the national value chain. The main objective of this project is to develop a rover to monitor coffee farms for rust detection. The robot consists of different subsystems: the first is to collect images to train the neural network, and the second is to monitor the coffee farms. The third subsystem, the Web subsystem, describes the challenges of wireless communication in coffee farms, detailing the characteristics of the hardware and the network configuration required to achieve such communication. The mechanical subsystem was developed based on a simulation model that was tested in different scenarios to ensure its operation in the coffee farm. It also describes the algorithms to mobilize the robot, including the detection of possible collisions and the proposed algorithm to avoid these collisions. Finally, the prototype and its limitations are presented to develop its work in coffee farms in Honduras.