Défense de mémoire en sciences informatiques : proactive computing and Machine Learning applied to advanced virtual robotics
Défense de mémoire de Christian Nazili, à distance via teams.
Date : 31/08/2020 08:30 - 31/08/2020 10:00
Lieu : https://teams.microsoft.com/l/meetup-join/19%3ameeting_ZmM2YWVlMTEtZWUzMC00MGQ3LThkZDktYTUwZjAyNTIyYzNi%40thread.v2/0?context=%7b%22Tid%22%3a%225f31c5b4-f2e8-4772-8dd6-f268037b1eca%22%2c%22Oid%22%3a%22363e7449-861f-4e7f-bbfa-9e6261d7a69a%22%7d
Orateur(s) : Christian Nazili
Organisateur(s) : Isabelle Daelman
This thesis proposes a solution to lighten the controllers of a robot by using machine learning. The goal is to reduce the complexity of the controller by entrusting some parts of the code to a machine learning algorithm.Theagentlearnsfromhisenvironmentthroughtherewardandpunishmentmechanism.Two learningalgorithmshavebeenusedandtested:Q-learningandSarsa.Theirimplementationwascarried out under the python programming language. The result of the algorithm is tested in a Webots robot simulator (with Thymbio II as robot). The open source mobile robot simulator, Webots, was chosen for this research. There are several choices of programming languages in Webots simulator (MatLab, Java, Python, C, C ++, ROS). For this work, the python language was used as the programming language.
Keywords : Reinforcement learning, robotic, Webots
Contact :
Isabelle Daelman
-
4966
-
isabelle.daelman@unamur.be
Télecharger :
vCal