TDDE13 Multiagent Systems

Examination

The examination consists of two parts:

1. LAB1 (2 ECTS)
   Complete the two lab assignments.
2. UPG1 (4 ECTS)
   Attend the two first seminars; solve and submit the home exercises before deadline; and write an individual report on a subject in multi-agent systems of your choice (a subject which you shall present at one of the two last seminars). Note that you need approval from the examinator or your TA on the subject you choose!

LAB1

LAB1 is now UPDATED
Please sign up in WebReg.

• You may work in groups of 2 or 3 students. Working alone is also allowed.
• Submit one report per group. Include all group members' names and LiU IDs.

There are two lab sessions in the course:

1. Agentic AI
   • Lab 1a: Building Your First LLM Agent
     Clone the lab repository from GitLab using your liuid https://gitlab.liu.se/amaso88/tdde13_assignment. Follow the instructions (lab1b/LAB1A_INSTRUCTIONS.md). The lab uses local open-source LLM models (Ollama with llama3.2) so no API keys required. To pass this lab, you need to complete all implementation tasks and submit your code and technical report.
   • Lab 1b: Multiagent Communication & Coordination
     Follow the instructions (lab1b/LAB1B_INSTRUCTIONS.md). Prerequisites: Completed Lab 1a with working single agent. To pass this lab, you need to complete all implementation tasks, and submit your technical report.
2. Multi-agent Reinforcement Learning Lab
   Download the instructions here.
   To pass this lab, you need your lab report approved by your TA.

UPG1

To pass UPG1, you need to complete the exercise sets and write an individual report. Each is graded separately. The grade on the exercises is given by the total points on the two assignments as follows:

Points	8-10	11-12	13-14
Grade	3	4	5

The total grade on UPG1 is given by the sum of the two grades diveded by 2 rounded down. You have to pass each part separately, i.e., get at least the grade 3.

Exercises	Report	Grade
3	3	3
3	4	3
4	3	3
4	4	4
3	5	4
4	5	4
5	3	4
5	4	4
5	5	5

Exercise Sets (UPG1)

There are two exercise sets in the course; one for each of the two first seminars:

1. Exercise Set 1 - Agents and Game Theory (7 points total)
   Download the instructions here.
2. Exercise Set 2 - Mechanism Design, Auctions and Social Choice (7 points total)
   Download the instructions here.

The exercise sets will be uploaded approximately two weeks before the corresponding seminar. See the timetable for deadlines. Follow the instructions in the exercise sets closely and use this LaTeX template (or this utf-8 version)! Also, be ready to answer questions about the problems at the corresponding seminar session.

Individual Report (UPG1)

To pass UPG1, you must also write an individual report on a subject related to multi-agent systems that you find interesting. You need to cite at least two published papers from well-known venues/journals. The report should be at maximum 3 pages (excluding references). It is important that you formulate yourself concisely and think carefully about what to include! You can choose any subject related to multi-agent systems, but here are a few suggestions (with sample papers for inspiration):

1. Cooperative Games, Coalition Formation, Coordination and Organizations
   Shehory, Onn, and Sarit Kraus. "Methods for task allocation via agent coalition formation." Artificial intelligence 101.1-2 (1998): 165-200.
   Sandholm, Tuomas, et al. "Coalition structure generation with worst case guarantees." Artificial Intelligence 111.1-2 (1999): 209-238.
   Ferber, Jacques, Olivier Gutknecht, and Fabien Michel. "From agents to organizations: an organizational view of multi-agent systems." International workshop on agent-oriented software engineering. Springer, Berlin, Heidelberg, 2003.
   Stone Peter, Gal A. Kaminka, Sarit Kraus and Jeffrey S Rosenschein "Ad hoc autonomous agent teams: Collaboration without pre-coordination." Twenty-Fourth AAAI Conference on Artificial Intelligence. 2010.
   Parag C. Pendharkar "Game theoretical applications for multi-agent systems" Elsevier Expert Systems with Applications Volume 39, Issue 1, January 2012
   Hannan Amoozad Mahdiraji, Elham Razghandi and Adel Hatami-Marbini. "Overlapping coalition formation in game theory: A state-of-the-art review" Elsevier Volume 174, 2021
   Khaled Abedrabboh, Luluwah Al-Fagih, Elena Parilina and Artem Sedakov. "Incentivising community self-consumption in energy markets: Stable coalition formation using cooperative game theory" Elsevier Applied Energy Volume 399, 2025
2. Multi-Agent Learning
   Lowe, R. et al. "Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments (MADDPG)." NeurIPS 2017
   Leibo, Joel Z., et al. "Multi-agent reinforcement learning in sequential social dilemmas." AAMAS 2017.
   Yu, Y. et al. "The Surprising Effectiveness of PPO in Cooperative Multi-Agent Games." NeurIPS 2022.
   Foerster, J. et al. "Counterfactual Multi-Agent Policy Gradients (COMA)." AAAI 2018.
   Le, H.M. et al. "Coordinated Multi-Agent Imitation Learning." ICML 2017
   Rashid, T. et al. "Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (QMIX)." ICML 2018.
   Zhengbang Zhu et al."MADIFF: Offline Multi-agent Learning with Diffusion Models." NeurIPS 2024.
   Hu, S. et al. "Modelling the Dynamics of Multiagent Q-Learning in Repeated Symmetric Games: A Mean Field Approach."N eurIPS 2019.
   Li, Y. et al. "Aligning Individual and Collective Objectives in Multi-Agent Mixed-Motive Games." NeurIPS 2024.
3. Non-Cooperative Games, Utility Theory & Equilibrium
   Nash, John. "Non-cooperative games." Annals of mathematics (1951): 286-295.
   Von Neumann, John, Oskar Morgenstern, and Harold William Kuhn. Theory of games and economic behavior (commemorative edition). Princeton university press, 2007.
   Fishburn, Peter C. "Retrospective on the utility theory of von Neumann and Morgenstern." Journal of Risk and Uncertainty 2.2 (1989): 127-157.
   Axtell, Robert L. "Non-cooperative dynamics of multi-agent teams." Proceedings of the first international joint conference on Autonomous agents and multiagent systems: part 3. ACM, 2002.
4. Agentic AI, LLM-based Agents and Multi-Agent Systems
   Shah, M. et al. "Large Language Model Based Multi-agents: A Survey of Progress and Future Directions." IJCAI 2024.
   Park, J.S. et al. "Generative Agents: Interactive Simulacra of Human Behavior." ACM 2023.
   Qian, C. et al. "ChatDev: Communicative Agents for Software Development" Arxiv
   Sen Yang et al. "Multi-LLM Collaborative Search for Complex Problem Solving" ICML 2024
   Kim, Y. et al. "MDAgents: An Adaptive Collaboration of LLMs for Medical Decision-Making." NeurIPS 2024
   Anjana Sarkar "Survey of LLM Agent Communication with MCP: A Software Design Pattern Centric Review" arxiv 2025

Students should propose their individual report topics by email. Each proposal should include:

1. A title
2. A short paragraph describing the focus of the study and how it relates to multi-agent systems. Example types of topics could be:
   • "Recent methods for multi-agent reinforcement learning in mixed cooperative-competitive environments"
   • "Comparison of coalition formation algorithms for task allocation"
   • "An overview of LLM-based multi-agent systems for collaborative problem solving"
3. A preliminary list of references (at least three papers).

Send in the suggestion to your TA for approval before starting to write your report. Use this LaTeX template for the final report.

Seminar Presentation (UPG1)

To pass the course, you also have to give a 10 minute presentation of your subject of choice (i.e., the one you are writing the individual report on) at one of the two last seminars (groups to be decided). This presentation should be aimed at your fellow students, so that they can understand and grasp the main ideas and concepts of your individual report.

Rules for examination of computer lab assignments at IDA

You are expected to do lab assignments in group or individually, as instructed for a course. However, examination is always based on individual performance.

It is not allowed to hand in solutions copied from other students, or from elsewhere, even if you make changes to the solutions. If there is suspicion of such, or any other form of cheating, teachers are obliged to report it to the University Disciplinary Board.

Be prepared to answer questions about details in specific code and its connection to theory. You may also be asked to explain why you have chosen a specific solution. This applies to all group members.

If you foresee problems meeting a deadline, contact your teacher. You can then get some help and maybe the deadline can be set to a later date. It is always better to discuss problems, instead of, e.g., to cheat.

Any kind of academic dishonesty, such as cheating (e.g., plagiarism, use of unauthorized assistance, and use of prohibited AI-based assistants) and failure to comply with university examination rules, may result in the filing of a complaint to the University Disciplinary Board. The potential penalties include suspension, warning.

Policy for handing in computer lab assignments at IDA

For all IDA courses having computer lab assignments there will be one deadline during or at the end of the course. If you fail to make the deadline, you must retake the, possibly new, lab course the next time the course is given.

If a course deviates from this policy, information will be given on the course web pages.

Page responsible: infomaster
Last updated: 2025-11-30