Goal-Capability-Commitment based Mediation for Multi-Agent Collaboration

Multi-agent collaboration in the agent oriented Collective Adaptive Systems could offset the failure of goal achievement by an independent agent. Commitments among agents to meld collaborative patterns typically defined in design time, which may cannot service the request of ACSs in response to changes in the system itself or in its environment. To address this problem, we propose a Goal-Capability-Commitment (GCC) based multi-agent collaboration mediation for ACSs at run-time. There are two issues with this state of the art approach: (1) the ability to represent the capability of agent under different contexts and diagnose the invalids of goals and agents leaded by the changing environments, and (2) the ability of mediation to select an appropriate capability or generate a new commitment in order to satisfy adaptive goal. We illustrate our approach through two experiments over the simulated scenario of automated hospital cart transportation system and discuss the results of experiments.