Multiagent systems 1000-2M00SW

Autonomous agents and multiagent systems or, more generally, multiagent environments represent a new approach to analysis, design and implementation of complex intelligent computer systems. An agent is a computational entity such as a computer program or a robot, capable of perceiving the environment and affecting it by performing actions. Usually it acts in cooperation with humans and/or other agents.

An agent is an autonomous intelligent entity. Its intelligence is reflected by the ability to reason and react in a flexible, rational way to unpredictable situations arising in the environment, as well as by the ability to take initiative, when necessary. As an interactive entity, an agent should be capable of working in heterogenous teams. In fact, the strength of multiagent systems, as compared to more conventional knowledge-base systems, lies in complex patterns of interaction. The paradigmatic example is a goal-directed cooperation, like teamwork, where a team works towards a complex goal that is beyond the individual reach.

Programme:

1. The notion of an agent and a multiagent system;
2. An overview of multiagent architectures:
- reactive architectures;
- hybrid architectures.
3. The BDI systems (i.e. systems consisting of agents defined in terms of beliefs, goals/desires and intentions).
4. The methods of solving conflicts in multiagent systems:
- auctions;
- negotiations.
5. Patterns of interaction in multiagent systems:
- distributed cooperative problem solving;
- methods of coordinating actions.
6. Communication in multiagent systems;

The course will be given in Polish, if no non-polish speaking students register for it.