Simulating drone automation in agriculture using neural networks

Mgr. Jakub Valent

Master's thesis

Abstract:

Táto práca sa snaží implementovať umelé neurónové siete do jednoduchých farmárskych procesov. Cieľom práce je použiť voľne dostupný softvér a knižnice na simuláciu scenára, v ktorom dron a kamerový senzor dokážu splniť jednoduchú úlohu zberu plodov paradajok pomocou neurónových sietí a reinforcment learning algoritmu. Pomocou algoritmov učenia zosilnenia sa nám podarilo naučiť dron nájsť, zobrať a priniesť cieľ na určené miesto v 3D priestore prostredia Unity. Pomocou L-Systemémov sa nám podarilo zkonštruovať 3D rastlinu paradajky s plodmi, na ktorej sme potom natrénovali neurónovú sieť.

Abstract:

This thesis tries to implement artificial neural networks into simple farming processes. The thesis aims to use freely available software and libraries to simulate a scenario in which a drone and a camera sensor can accomplish the simple task of harvesting a tomato fruit using neural networks and reinforcement learning. Using reinforcement learning algorithms, we were able to teach a drone to find, pick up and bring the target to a designated location in the 3D space of the Unity engine environment using the camera. Using L-Systems, we were able to construct a 3D tomato plant with fruits, on which we then trained a neural network. iv

Keywords

drone, neural, network, artificial intelligence, agriculture, Unity 3D, simulation, L-System

Language used: English

Date on which the thesis was submitted / produced: 13. 12. 2022

Thesis defence

• Date of defence: 27. 1. 2023
• Supervisor: doc. RNDr. Tomáš Brázdil, Ph.D.
• Reader: RNDr. Petr Novotný, Ph.D.

Masaryk University

Faculty of Informatics

Master programme / field:
Visual informatics / Computer Games Development