Evaluation of Trust Algorithms in Competitive Multi- Agent Systems

DOI: 10.5176/978-981-08-7466-7_se-52

Authors: Michael deFreitas, Mansour Z and Harvey Siy

Abstract:

In any multi-agent system (MAS) where agents mustrely on each other in order to achieve individual or shared goals,the issue of trust is important. How does one agent decide whether or not to rely on another, particular agent to perform at ask? How does the agent record the results of any interaction, and possibly use that record in future decisions? How should reputation information be shared between agents, and what level of trust should be placed in that information? These questions are at the forefront of recent research into trust and reputation in MAS, and most research focuses on knowledge representation and algorithms for interpreting that knowledge when making trust decisions. One area not deeply explored is the effect of the MAS environment itself on trust decisions. How should each type of environment – or even an environment where the level of competitiveness changes over time – affect a participating agent’s trust decisions? This work presents design and implementation of a test bed for a trust algorithm that explores environmental factors in an agent’s trust. The main objective of this paper is to represent the details of implementation aspect of this project.

Keywords: multi-agent systems; trust computation