Algorithms for automated multilateral negotiations

Αλγόριθμοι για αυτοματοποιημένες πολυμερείς διαπραγματεύσεις (EL)

Algorithms for automated multilateral negotiations (EN)

Ξενου Κωνσταντια (EL)

Xenou Konstantia (EN)

Λαγουδακης Μιχαηλ (EL)

Χαλκιαδακης Γεωργιος (EL)

Πολυτεχνείο Κρήτης (EL)

Σπανουδακης Νικολαος (EL)

Chalkiadakis Georgios (EN)

Lagoudakis Michael (EN)

Technical University of Crete (EN)

Spanoudakis Nikolaos (EN)

In multi-attribute negotiations, two or more intelligent automated agents sharing common or conflicting interests negotiate over various distinct issues, usually under uncertainty about the characteristics of their negotiating partners. Agents should be able to adaptively adjust their behavior during the negotiation process, by adopting efficient and effective automated negotiation techniques. This is a challenging process, since negotiators are not usually willing to reveal their private information and preferences, so as to avoid being exploited by the other participants during the negotiation. To overcome these problems and boost agent performance, a modeling of opponents’ preferences and strategy is usually incorporated—with the understanding that uncertainty regarding opponent preferences always exists in real-world settings. Opponent modeling can be usually shown to assist the agent to achieve efficient agreements, and thus to significantly increase the quality of the negotiation outcome. A multitude of negotiation strategies and opponent models have been coined and studied over the years; regardless, empirically comparing them to each other is not a straightforward exercise. To this end, the international Autonomous Negotiation Agents Competition (ANAC) was initiated in 2009, and is conducted utilizing a “standard”, purpose-built, negotiation platform (“Genius”). Genius provides a uniform, accepted by all, way of comparing state-of-the-art agent strategies. In this thesis, we systematically developed several different negotiating strategies, along with ac-companying opponent models. We employed concepts found in the literature, implemented known strategies, and proposed novel ones. This process resulted to the creation of thirteen (13) distinct agents. The developed agents were pitted against previous ANAC participants, and also against each other, with an extensive evaluation being conducted on Genius. One of our strategies was se-lected and participated in the international ANAC-2017 competition. Our thesis presents a detailed evaluation and analysis of the performance of all our agents. (EN)

bachelorThesis

Multiagent systems (EN)

Negotiations (EN)

Technical University of Crete

Institutional Repository Technical University of Crete

School of Electrical and Computer Engineering - Master Theses

English

2017

http://purl.tuc.gr/dl/dias/EEF4F4F7-49C6-4FF7-9190-F275E332088D

Πολυτεχνείο Κρήτης::Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών (EL)

Technical University of Crete::School of Electrical and Computer Engineering (EN)