Improvements in performance of large-scale multi-agent systems based on the adaptive/non-adaptive agent selection

An intelligent agent in a multi-agent system (MAS) often has to select appropriate agents to assign tasks that cannot be executed locally. These collaborating agents are usually determined based on their skills, abilities, and specialties. However, a more efficient agent is preferable if multiple candidate agents still remain This efficiency is affected by agents' workloads and CPU performance as well as the available communication bandwidth. Unfortunately, as no agent in an open environment such as the Internet can obtain these data from any of the other agents, this selection must be done according to the available local information about the other known agents. However, this information is limited and usually uncertain. Agents' states may also change over time, so the selection strategy must be adaptive to some extent. We investigated how the overall performance of MAS would change under adaptive strategies. We particularly focused on mutual interference by selection in different workloads, that is, underloaded, near-critial and overloaded stituations. This paper presents the simulation results and shows the overall performance of MAS highly depends on the workloads. Then we explain how adaptive strategies degrade overall performance when agents' workloads are near the limit of theoretical total capabilities.