A Foundational Analysis of Software Robustness Using Redundant Agent Collaboration

This paper describes an investigation into how the robustness of a software system can be improved via the use of software agents as a fundamental component of the system. Robustness can be achieved through redundancy, and we hypothesize that agents are an appropriate unit for adding redundancy. As part of our investigation, we asked 53 students to independently build one agent for a simple Multiagent domain. We did a comparison study between the different student’s agents while working with each other independently and while working with each other as part of collaboration. A simple decision-making mechanism was also introduced and compared. This influx of redundancy in the collaborations provided a discernible increase in productivity. The results shown here should lead to further rich and enlightening research into agent-oriented software engineering.