Negotiation strategy for multilateral multi-issue negotiation

• Main Author: Neo, Harry Jun Guo
• Other Authors: Bo An
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/62816
• Institution: Nanyang Technological University

Negotiations have become an essential part of our daily lives, giving rise to numerous applications adopting the use of automated negotiation systems. The capabilities that such systems possess are endless and have begun to adapt towards being able to serve more complex tasks. This has compelled research to focus in this discipline. The Automated Negotiation Agents Competition (ANAC) is one such initiative that has provided a unique benchmark to evaluate negotiation strategies that have been developed over the years. In this project, a negotiation agent is developed and implemented in accordance with the rules and regulations of the ANAC 2015. The agent was developed while taking into account the various considerations spelt out i.e. the need for opponent modelling, the optimal acceptance threshold etc. The agent adopted an approach that enlisted the usage of tokens of endorsements and an incremental frequency-learning based approach. This allowed the agent to subtly create an alliance within a negotiation and make its offer more acceptable to opponents. The agent strategy was able to place second in the tournament simulations where it was pit against 3 other agents in linear scenarios. The agent handled itself well against both tough and soft opponents, achieving consistent results with each negotiation session. In the future, more negotiation agents could be included in the evaluation of the agent developed as there are limited choices at the present moment given that this is the first time the ANAC is conducting a multi-lateral negotiation tournament.