Towards a Definition of Robustness for Market-Style Open Multi-Agent Systems

摘要

Multi-agent systems (MAS) are an integral part of current (D)AI research. Their success and the success of agent technology is in the majority of the literature ascribed to three major advantages, one of which is that multi-agent systems are robust systems (cf. [1, 9]). Although the literature treats robustness like an inherent feature of MAS (which it is not), there is hardly any discussion on what robustness actually means (a rare exception is the work of Kaminka and Tambe and Klein and Dellarocas). Just like any other artificial complex system, multi-agent systems need to be specifically designed to be robust. Compared to conventional computer science, the issue of robustness in MAS is different. Most computer science systems are transformational systems, which means they compute a function on some input. Here, techniques for ensuring robustness exist (e.g. cf. [6]). However, the most interesting multi-agent systems are open systems, which do not explicitly compute a function (e.g. looking at the most predominant example of an open system, namely the Internet, as computing a function certainly misses the point).