Veranstaltung
Die hier angezeigten Termine und Veranstaltungen werden dynamisch aus Stud.IP heraus angezeigt.
Daher kontaktieren Sie bei Fragen bitte direkt die Person, die unter dem Punkt Lehrende/Dozierende steht.
Veranstaltung
Semester:
Wintersemester
2023
2.01.591 Verification of Distributed AI Systems -
Veranstaltungstermin | Raum
- Mittwoch, 18.10.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 25.10.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 1.11.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 8.11.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 15.11.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 22.11.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 29.11.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 6.12.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 13.12.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 20.12.2023 10:00 - 12:00 | A05 2-229
- Mittwoch, 10.1.2024 10:00 - 12:00 | A05 2-229
- Mittwoch, 17.1.2024 10:00 - 12:00 | A05 2-229
- Mittwoch, 24.1.2024 10:00 - 12:00 | A05 2-229
- Mittwoch, 31.1.2024 10:00 - 12:00 | A05 2-229
Beschreibung
This module is a reading group seminar of current research on verification techniques for multi-agent systems (MAS).
Agents are (partially) independent units with the ability to observe their local environment, communicate with other agents and make decisions.
MAS are a highly flexible paradigm making them scalable, adaptable and resilient.
This comes at the cost of increased complexity, making hard guarantees on the behavior of MAS difficult to prove, especially when emergent system behaviors can not be easily derived from individual agent actions.
In this seminar we will read and discuss recent research on ways to deal with this complexity. These techniques can apply to different levels of conceptualization, including (but not limited to):
* describing the MAS as a whole:
The paper selection is mostly decided by the students within the bounds of the seminar topic.
Active participation in these discussions is the basis for evaluation.
Agents are (partially) independent units with the ability to observe their local environment, communicate with other agents and make decisions.
MAS are a highly flexible paradigm making them scalable, adaptable and resilient.
This comes at the cost of increased complexity, making hard guarantees on the behavior of MAS difficult to prove, especially when emergent system behaviors can not be easily derived from individual agent actions.
* describing the MAS as a whole:
* interpreted systems
* algorithmic view of agent actions* automata
* emergence design (guaranteeing system level properties by design)
* emergence design (guaranteeing system level properties by design)
* distributed optimization algorithms
* local agent view* behavioral verification of individual agents and consequences for system properties
The seminar will consist of 2 week cycles. We decide on a paper and discuss it 2 weeks later in the next meeting. The paper selection is mostly decided by the students within the bounds of the seminar topic.
Lehrende
SWS
2
Lehrsprache
deutsch und englisch