On Manageability and Robustness of Open Multi-agent Systems

This paper addresses the manageability of large, open and distributed multi-agent systems (MASs). Specifically, we are concerned here with three critical elements of management: (a) the ability of the manager to monitor relevant operations of its subordinates; (b) its ability to steer its subordinates; and (c) and the robustness of the agentcommunity under certain unexpected adverse conditions, such as unpredictable failure of the manager itself. Using our own law-governed interaction (LGI) mechanism, we show how one can build a scalable infrastructure that supports monitoring, steering and a degree of robustness, in spite of the heterogeneous and dynamic nature of the system being managed, and in spite of possible changes and failures of the manager itself.