Opponent Modeling with Information Adaptation (OMIA) in Automated Negotiations
RIS ID
117958
Abstract
Springer International Publishing AG. 2017. Opponent modeling is an important technique in automated negotiations. Many of the existing opponent modeling methods are focusing on predicting the opponent's private information to improve the agent's benefits. However, these modeling methods overlook an ability to improve the negotiation outcomes by adapting to different types of private information about the opponent when they are available beforehand. This availability may be provided by some prediction algorithms, or be prior knowledge of the agent. In this paper, we name the above ability as Information Adaptation, and propose a novel Opponent Modeling method with Information Adaptation (OMIA). Specifically, the future concessions of the opponent will firstly be learned based on the opponent's historical offers. Then, an expected utility calculation function is introduced to adaptively guide the agent's negotiation strategy by considering the availability and value of the opponent's private information. The experimental results show that OMIA can adapt to different types of information, helping the agent reach agreements with the opponent and achieve higher utility values comparing to those which lack the information adaptation ability.
Grant Number
ARC/DP140100007
Publication Details
Wang, Y., Ren, F. & Zhang, M. (2017). Opponent Modeling with Information Adaptation (OMIA) in Automated Negotiations. Lecture Notes in Computer Science, 10642 21-35. International Conference on Autonomous Agents and Multiagent Systems AAMAS 2017: Autonomous Agents and Multiagent System