Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Introducing Machine Learning Using Robots - Design and Integration of Simple Neural Networks and the Q-learning Algorithm in the Robot Simulation Environment of Open Roberta Lab, Accompanied by the Development, Testing, and Evaluation of Complementary Teaching Materials

: Olari, Viktoriya
: Eide, Øyvind

Volltext urn:nbn:de:0011-n-6384972 (130 MByte PDF)
MD5 Fingerprint: a89cabf235b56a403398f9f3d862623f
Erstellt am: 3.8.2021

Köln, 2020, XI, 191 S.
Köln, Univ., Master Thesis, 2020
Master Thesis, Elektronische Publikation
Fraunhofer IAIS ()
AI Education; AI Teaching; AI Learning; ML Education; ML Teaching; artificial intelligence; machine learning; robotics; playful learning; neural networks; Q-Learning; clustering; K-means Algorithm; Open Roberta Lab; Constructionism

The following masters thesis provides an approach to introducing machine learning to students using the block-based programming language NEPO in combination with educational robotics. The target group of the research study are students from primary to high school, representing beginners without any previous knowledge of machine learning. After analysing the guidelines and methods for the introduction of machine learning in schools, as well as concrete proposals for artificial intelligence school curricula, the author identified a large discrepancy between the requirements for introducing the topics of supervised, unsupervised, and reinforcement learning in schools and the solutions currently available on the educational landscape to do so. Most of the approaches which are currently available either remain a black box or are inaccessible to young students. In order to close this discrepancy, and following the ideas of constructionism, the author developed three approaches to introduce machine learning using robots: (1) The Neural Network Playground which allows the user to experiment with simple neural networks, (2) The Q-learning Playground which enables the student to tinker with the Q-learning algorithm, (3) An unplugged activity introducing the k-means algorithm that makes the unsupervised learning tangible. The author accompanied all approaches with a curriculum and a series of learning materials. She then conducted and evaluated a user study with 24 children from primary, middle, and high school. The results underline the practical feasibility of the approaches: the children of all age groups perceived the topics as interesting and ranging from very easy to moderately hard to grasp. Thus, the research study proposes a solid concept for the introduction of machine learning to beginners which fundamentally differs from the currently available approaches and enriches the educational landscape.