Recently, commercial drones have become very affordable and have been used in many fields and applications. In particular, farm drones have been used as substitutes for human workers in agriculture. The thesis aims to create a simulation of drone flight control and a garden of tomato plants with growth to simulate a farm drone. Moreover, the drone should be controlled using deep learning methods trained in the simulator. Concretely, the drone should be trained to recognize the states of the tomato plants and their fruits and pick up the fruits when ripe. The text should also describe related work on learning drones in agriculture.