Overview

This is an advanced semiar on (cooperative) multi-agent reinforcement learning (MARL). Some knowledge of (single-agent) reinforcement learning is assumed and we will cover the principles and state-of-the-art methods in MARL. We will focus on the cooperative (team) MARL setting but also discuss how these methods apply to the more general setting.

Readings will be from:

Some introductory material from the (much simpler) 4100/5100 class:  intro to MDPs, policy iteration, online plannig, Intro to RL simple model-based RL, intro to model-free RL, Q-learning, exploration, generalization, intro to policy gradient

Schedule

Date Topic Notes Reading Assignment out/due
9/4 Introduction pptx
9/8 RL Review 1 pptx, zoom link (sorry about the quality!) SB:  3.1--3.8, 4.1--4.8, 5.1--5.7
9/11 RL Review 2 pptx, zoom link
 SB: 6.1--6.8, 8.1--8.6;8.9--8.12, 9.1--9.5, 9.8, 10.1, 13.1--13.7  
9/15 RL Review 3 pptx, zoom link    
9/18 Partially Observable RL and Centralized MARL    
9/22 Centralized MARL?      
9/25 CTDE      
9/29 More CTDE
 
10/2 Paper presentations      
10/6 Paper presentations

 
10/9 Paper presentations

 
10/13 No class (Indigenous Peoples Day)

10/16 Paper presentations    
10/20 Paper presentations     
10/23 Paper presentations
   
10/27  Paper presentations    
10/30 Paper presentations
 
11/3 Paper presentations
11/6 Paper presentations  
11/10 Paper presentations
11/13 Paper presentations
11/17 Paper presentations


11/20 Paper presentations

11/24 Paper presentations


11/27
 No class (Thanksgiving)


12/1
 Project Presentations

 
12/4 Project Presentations


12/8 Project Presentations


   


12/10
 Project Reports Due

Report due at 11:59 PM -- This is a hard deadline, no extensions


Important note: unless noted otherwise, all readings and assignments are due on the day that they appear in the schedule.