Hauptinhalt

Topinformationen

Begleitseminar zum Studienprojekt EBiMAS: Emergent behaviors in a multi-agent system with reinforcement learning (Part V)

8.3579

Dozenten

Beschreibung

In the EBiMAS study project, you will collaborate as a team on developing and training a multi-agent system with a 3D simulation framework that we’ve developed based on Google DeepMind’s MuJoCo physics engine. In the scenario you will work on, simple artificial ant agents must solve a puzzle-foraging task, potentially learning to communicate with each other in the process. The agents are trained with advanced Deep Reinforcement Learning algorithms to investigate emergent properties and dynamics between the agents, such as language, cooperation, and strategies.

Prerequisites for this course include proficiency in programming (Python) and either basic experience with simulation tools like Blender, MuJoCo, Nvidia Isaac, or foundational knowledge of Machine Learning (ML) / Reinforcement Learning (RL). Applicants for the course are asked to submit a brief essay before the start of the semester, introducing themselves, their background, relevant courses and experiences, as well as highlighting their motivation for participating in the project (mail as PDF to julius.mayer@uni-osnabrueck.de). This will help us better understand each candidate and determine the best fit for the program.

Full-team milestone meetings with the supervisors are held bi-weekly on Wednesdays. Specific times may be adjusted based on group availability. Students are expected to self-organize as a team 2-3 times per week, with meeting times for group work agreed upon among students and subgroups.

Note: This course is open to students starting their first part of a two-semester study project in WS24 and those continuing from the previous semester. There is also a separate one-semester interdisciplinary seminar that accompanies the study project.

Weitere Angaben

Ort: nicht angegeben
Zeiten: Mi. 16:30 - 18:30 (wöchentlich) - There will be bi-weekly team meetings for updates and planning with supervisors. Students are expected to self-organize and meet regularily in their work groups 2-3 times per week. Meeting times may adjust based on group availability.
Erster Termin: Mittwoch, 30.10.2024 16:30 - 18:30
Veranstaltungsart: Seminar (Offizielle Lehrveranstaltungen)