|Adm. – Grad.||2017 –|
|Dir.; Codir.||Stéphane Gagnon|
Integrating Semantic Reasoning in a Multi-Agent System in Erlang
Context: Since the revolution of Object-Oriented Programming (OOP), and especially since about 2005, software is undergoing its next revolution towards explicit concurrent and parallel programming due to the industry major change to the parallel hardware. As a result, parallel, heterogenous, and distributed hardware and systems have become the mainstream available universally in personal computers, smaller devices, and networked applications. This fundamental parallel hardware shift requires seeking the best ways to address the challenges of explicit parallel programing and properly utilize multicore and distributed CPUs. Multi-Agent Systems (MAS), consisting of agents that operate and cooperate concurrently, are traditionally modeled based on OOP concepts, mainly Java programming language as a mainstream of OOP.
Problem: However, the nature of MAS and their fundamental assumptions contradict with OOP principles, especially in terms of OOP shared memory model, and its inherited complexity of concurrent programming as well as its lack of native support for intelligent aspects of agents. Recent trends show that functional programming paradigm provides promising strategies of abstraction and features specifically meeting the requirements and addressing the complexity of concurrent or parallel programming including MAS features. Among several function programming languages, Erlang functional programming language provides one of the most straightforward and inexpensive approaches to exploit distributed and multicore CPUs concurrently efficiently as well as to provide all-in-one environment to program all aspects of MAS.
Methodology: This research proposes to reuse and extended the capabilities of 2 open source Erlang libraries, erlang eXperimental Agent Tool (eXAT) and Swarm oriented ERlang Expert SYstem Engine (SERESYE), by integrating semantic web knowledge models and standards such as Resource Description Framework (RDF) data structures and Web Ontology language (OWL) ontologies in a rule engine that allows agents to perform various tasks of reasoning and analysis over such emerging knowledge models and standards. We propose to demonstrate well-known use cases, such as the Book Trading case of Java Agent DEvelopment (JADE). Moreover, the system will be made available via a web portal for users testing and feedback. The feasibility and efficiency of the system will be evaluated based on the recall and precision of the rules-based scenarios, vs. actual optimal decisions, and performance metrics including use of resources, time, and scalability.
Outcome: This demonstration of how we can integrate semantic reasoning in a multi-agent system will allow us to develop more complex and intelligent collaborative systems. The applications will be numerous, primarily in Robotic Process Automation (RPA) in Industrie 4.0, and other rule-based and formally defined problems.