Petri-net integration – An approach to support multi-agent process mining

During the developing phase of a multi-agent system (MAS), it is necessary to simulate the system feasibility before the formal implementation. Petri-net (PN) is the most popular tool used in MAS system simulation. However, most of the PN models were developed according to the investigation of the MAS behaviors. There lacked a bridge to directly link MAS development tool and the PN simulator. This research proposed a PN integration approach to support multi-agent process mining. A process mining technique (α algorithm) would be used to investigate the log file generated by each agent system and then to create the PN based process models. In the next stage, an approach was proposed to integrate the PN of each agent. This method would construct the incoming and outgoing places of each agent’s PN according to inter-agent communication behaviors showed in the logs. These incoming and outgoing places would be merged into communication places representing the inter-relationship between agent’s PN models. Finally, the multi-agent PN model would be simulated and the results showed that the agent interaction behaviors are consistent with the information shown in their log files.