Achieving software robustness via large-scale multiagent systems

This paper describes how multiagent systems can be used to achieve robust software, one of the major goals of,software engineering. The paper first positions itself within the software engineering domain. It then develops the hypothesis that robust software,can be achieved through redundancy, where the redundancy is achieved by agents that have different algorithms but similar responsibilities. The agents are produced by wrapping conventional algorithms with a minimal set of agent capabilities, which we specify. We describe our initial experiments in verifying our hypothesis and present results that show an improvement in robustness due to redundancy. We conclude by speculating on the implications of multiagent-based redundancy for software development.