USE OF MACHINE LEARNING FOR TRAINING A STRAWBERRY SPRAYING ROBOT IN A UNITY 3D ENVIRONMENT

Authors

  • Timotej Orčić Autor

DOI:

https://doi.org/10.24867/10BE20Orcic

Keywords:

artificial intelligence, artificial neural net¬works, object detection, Reinforcement Learning, Unity 3D, Unity ML Agents

Abstract

The aim of this research is to present a conceptual solution for rapid development of artificially-intelligent agents through simulation in a Unity graphical environment. The area of application of the specifically developed agent is agriculture. An object detection model was trained, as well as a Reinforcement Learning model for targeted agent movement in the environment. Data and the environment are self-created. All models and algorithms used were evaluated.

References

Svim navedenim linkovima je pristupljenu u junu 2020.
[1] https://unity.com/
[2] A. Juliani, V-P. Berges, E. Teng, A. Cohen, J. Harper, C. Elion, C. Goy, Y. Gao, H. Henry, M. Mattar, D. Lange. Unity: A General Platform for Intelligent Agents, 2018
[3] https://github.com/tensorflow/models/tree/master/-research/object_detection
[4] https://colab.research.google.com/notebooks/-intro.ipynb
[5] Mnih, V., Badia, A.P., Mirza, M., Graves, A., Lillicrap, T., Harley, T., Silver, D. and Kavukcuoglu, K., 2016, June. Asynchronous methods for deep reinforcement learning. In International conference on machine learning (pp. 1928-1937).
[6] https://www.tensorflow.org/
[7] http://wiki.fast.ai/index.php/Fine_tuning
[8] http://cocodataset.org/

Published

2020-10-31

Issue

Section

Electrotechnical and Computer Engineering