Runtime Verification of Multi-agent Systems Interaction Quality

摘要

Since multi-agent systems are inherently complex, there are possibilities that errors related to multi-agent systems interaction could occur. Currently, many verification approaches have been proposed by focusing on specific properties, using a particular technique and during certain development phase. However, each technique has its limitations. As interaction between agents and multi-agent systems environments evolve during runtime, not all multi-agent systems interaction requirements can be specified and verified during design and development. Thus, some new interaction properties such as agent availability and trustability need to be verified during runtime. In this research, a solution is proposed in which newly defined agents interaction quality requirements are specified, developed into metrics and verified within multi-agent systems runtime verification framework. It is aimed to improve the effectiveness of the verification of agent interactions during runtime. Finally, an experiment is set up to capture message passing between agents and to gather runtime system profiles to evaluate the proposed solution.