Agents and Multiagent Systems

• 26 hrs lectures
• 13 hrs pc labs
• 13 hrs projects

• 60 pts final exam (written part)
• 20 pts mid-term test (written part)
• 20 pts projects

The objective of the course is to introduce students to the operational principles and design of agent-based systems featuring autonomous intelligent units  as well as multi-agent systems composed of multiple interacting agents. Students will learn to design, model, and program these systems and their individual components using both declarative and asynchronous modern software frameworks.

Graduates will acquire comprehensive theoretical and practical knowledge of contemporary approaches to building autonomous intelligent entities capable of operating in highly dynamic environments. They will understand the formal foundations of agent architectures, the BDI model, and FIPA standards for agent communication. Furthermore, students will gain deep insight into multi-agent problem-solving, including negotiation, conflict resolution, coalition formation, and distributed planning.

Practical lab sessions provide hands-on experience in implementing intelligent agents and multi-agent systems using Jason, JADE, and SPADE in Python. Students will also explore the integration of classical agent architectures with Large Language Models (LLMs) and the fundamentals of Multi-Agent Reinforcement Learning.

It is necessary to have fundamental knowledge of formal logic, artificial intelligence, system modelling and programming for this course.

• Russel, S., Norvig, P.: Artificial Intelligence, a Modern Approach, Pearson Education Inc., 2003, ISBN 0-13-080302-2
• Wooldridge, M.: Reasoning about Rational Agents, 2000, The MIT Press, Cambridge, MA, ISBN 0-262-23213-8
• Shoham, Y, Leyton-Brown, K.: Multiagent systems, Algorithmic, Game-Theoretic, and Logical Foundations,  Cambridge University Press, 2009
• Shoham, Y, Leyton-Brown, K.: Multiagent systems, Algorithmic, Game-Theoretic, and Logical Foundations,  Cambridge University Press, 2009

• Russel, S.Norvig, P.: Artificial Intelligence, a Modern Approach, Pearson Education Inc., 2003, ISBN 0-13-080302-2 

• Wooldridge, M.: Reasoning about Rational Agents, 2000, The MIT Press, Cambridge, MA, ISBN 0-262-23213-8

• Woolridge, M.: Introduction to Multiagent Systems, Wiley, 2009

1. Introduction to Distributed Artificial Intelligence: Concept of an agent and multi-agent systems (MAS), environmental properties, agent typologies, and classifications.

2. Formal Agent Models and Reactive Architectures: Wooldridge’s formal model of agents and environments, perceptions, states, and behavior functions (demonstrated on TileWorld). Reactive approach, situated automata, subsumption architecture, and layered architectures.

3. Formal Logics in Agent Systems: Modal logics, possible-world semantics, Kripke structures, temporal logics (LTL and CTL), formal foundations of BDI logic.

4. Mental States and BDI Architectures: Rational agents, intent-driven systems and their modeling, IRMA architecture, AgentSpeak framework.

5. Advanced BDI Approaches: Extensions of the BDI model, intent synergies, late variable binding during plan execution to increase reasoning flexibility.

6. BDI Programming in Jason: Practical aspects of BDI agents, environment interfaces, and architecture of the Jason system; demonstrations and code walkthroughs in AgentSpeak(L).

7. FIPA Standards, Communication, and MAS Engineering: Abstract architecture and agent lifecycle, KQML and FIPA ACL languages, interaction protocols. GAIA and Prometheus design methodologies, agent roles in MAS.

8. MAS Implementation Platforms and LLM Agents: Principles and architecture of JADE, transitioning to Python via the SPADE framework. Utilizing Large Language Models (LLMs) for agent reasoning, tool calling, Model Context Protocol (MCP), and LLM integration into SPADE.

9. Negotiation, Voting, and Argumentation in MAS: Physical and mental conflicts, game theory, Rubinstein’s bargaining model, voting mechanisms, and formal argumentation.

10. Coalitions and Coalitional Games: Theory of coalitional games, solution concepts (Core, Shapley value), and algorithms for coalitional structure generation (DP, Anytime IP).

11. Coordinated Action and Distributed Planning: Centralized vs. distributed planning, intent coordination methods, and Distributed Constraint Satisfaction/Optimization Problems (DisCSP/DCOP).

12. Social Behavior in MAS: Principles of agent societies, concepts of commitment, norms, and social structures.

13. Learning in Agent Systems: Reinforcement Learning (RL) and Multi-Agent Reinforcement Learning (MARL) for agent behavior.

1. Two practices - JASON system
2. Two practices - JADE system
3. Two practices - JACK system

Team project - design of a mulitagent system, cooperative planning, coordination, negotiation

• Mid-Term test
• Team project
• In order to obtain points in the exam, the exam must be prepared in such a way that it is scored higher than the minimum threshold indicated in the information system (25 points). Otherwise, the exam will be scored 0 points.