A Grammatical Approach for Distributed Business Process Management using Structured and Cooperatively Edited Mobile Artifacts

TL;DR

This thesis presents a novel distributed workflow system for automating administrative business processes, utilizing multiagent, P2P, SOA architectures, and cooperative editing of structured artifacts to enable decentralized, confidential task execution.

Contribution

It introduces a grammatical model for process specification integrated into a P2P system, enabling autonomous, decentralized execution of workflows with cooperative artifact editing.

Findings

Mathematical tools for expressing workflows as attributed grammars.

Decentralized execution of processes via autonomous agents and artifacts.

Support for confidential tasks with partial process perception.

Abstract

In this thesis, we focus on the proposal of distributed workflow systems dedicated to the automation of administrative business processes. We propose an approach to build such systems by relying on the concepts of multiagent systems, Peer to Peer (P2P) architecture, Service-Oriented Architecture (SOA) and structured documents (artifacts) cooperative edition. Indeed, we develop mathematical tools that allow any workflow systems designer, to express each administrative process in the form of an attributed grammar whose symbols represent tasks to be executed, productions specify a scheduling of these tasks, and instances (the derivation trees that conform to it) represent the different execution scenarios leading to business goal states. The obtained grammatical model is then introduced into a proposed P2P system which is in charge of carrying out the completely decentralised execution of the underlying process's instances. The said system orchestrates a process's instance execution as a choreography during which, various software agents driven by human agents (actors), coordinate themselves through artifacts that they collectively edit. The exchanged artifacts represent the system's memory: they provide information on already executed tasks, on those ready to be executed and on their executors. The software agents are autonomous and identical: they execute the same unique protocol each time they receive an artifact. This protocol allows them to identify the tasks they must immediately execute, to execute them, to update the artifact and to disseminate it if necessary, for the continuation of the execution. Moreover, actors potentially have only a partial perception of processes in which they are involved. In practice, this means that certain tasks can be carried out confidentially.