Аспирантский семинар: Discovering Process Models for Multi-Agent Systems from Event Logs

Where: Faculty of Computer Science, Kochnovskii proezd, 3, room 205.
When: 18:10 – 19:30 on December 3.
Speaker: Roman Nesterov.

In this talk, I will present a modular approach to discover workflow nets for multi-agent systems from event logs. Many process discovery algorithms have been proposed over recent years. However, multi-agent system models, directly synthesized by existing algorithms, can be incomprehensible since concurrently interacting agents produce rather sophisticated behavior. To overcome this problem, a compositional technique is proposed. Event logs of multi-agent systems are filtered according to an individual agent behavior. We discover a workflow net for each agent using existing tools. This study considers only asynchronously interacting agents. This is formalized in terms of a channel-composition operation. Given a specification of an interaction protocol, we provide a general scheme for composing workflow nets. As for the main result, we prove that, by applying morphisms, this composition preserves soundness of agents within a model for a whole system.