Approaches to implementing decision aggregation in multi-agent systems

Decision aggregation is about the combination of different decisions from different decision makers to obtain an overall decision. It is an important issue in multi-agent systems (MASs) because in many multiagent application systems, each agent may only have a limited amount of domain knowledge or information. An agent can make decisions based on its existing knowledge, however, its decisions will have to be analyzed and combined with other agents' decisions. Researchers have invested significant resources in devising and evaluating a large array of aggregation algorithms, but few involved in the implementations of these aggregation algorithms in MASs. In this paper, we propose three approaches to implement decision aggregation in MASs- by using specific stationary aggregation agent, by utilizing token passing mechanism, and by employing mobile aggregation agent. Their implementation details are presented. The advantages and disadvantages of each approach are also discussed. After comparing these three approaches, we argue that the first two are better in small-scale MASs, but the mobile agent approach is most promising and practical in large-scale MASs.