Multiagent Reinforcement Learning (seminar)

Course information

• Organized by:  Prof. Joschka Boedecker, Shengchao Yan
• Course number on HISinOne: 11LE13S-7368-M
• Introductory lecture: Friday, April 24, 15:00-16:00, NEXUS Lab, IMBIT building (Georges-Köhler-Allee 201)
• Regular location: 

      IMBIT (Georges-Köhler-Allee 201)
      Room 42 (1st floor)

• Regular time: Wednesday 14:00 - 15:30
• Language: English

Overview

Multi-agent reinforcement learning (MARL) sits at the heart of many of today’s most exciting AI challenges: it studies how multiple decision-making entities learn and adapt in shared environments where the outcome for each depends on the actions of others. From autonomous vehicles negotiating traffic, to decentralized energy grids, financial markets, and large-scale online platforms, real-world systems are inherently multi-agent and often require a balance of cooperation and competition. Unlike single-agent learning, MARL introduces fundamentally new difficulties, non-stationarity, strategic behavior, and coordination, making it both theoretically rich and practically impactful.

This seminar explores the rapidly growing field of MARL, guided by selected chapters from the book Multi-Agent Reinforcement Learning: Foundations and Modern Approaches. The focus is on understanding how multiple learning agents interact, cooperate, and compete within shared environments.

Format

The seminar will rely on in-person attendance. After a short ramp-up phase with an introductory lecture, we will start with presentations and discussions on chapters of the book "Multi-Agent Reinforcement Learning: Foundations and Modern Approaches". 

Students will work in pairs, each group taking responsibility for one chapter. Your task is to:

• Read and understand the assigned material
• Present the key ideas, methods, and results
• Lead a discussion with your peers

Important Note on Prerequisites

This seminar largely focuses on modern approaches of MARL, which is based on the recent developement of deep learning and reinforcement learning. Therefore, this seminar accepts students who have completed graded courses equivalent to Foundations of Deep Learning and Reinforcement Learning.

IMPORTANT: For this semester, we can only accept a limited number of students for this course. During the first session, we will clarify which prerequisites are required, and we will ask students without appropriate prerequisites to de-register from the course.

Materials

Presentation templates: either the university template or the latex template provided by our lab.

Grading criteria: your participation will be graded according to these guidelines [to be added].

Timeline

Questions?

If you have a question, please post it in the ILIAS forum (so everyone can benefit from the answer).