Marc-Philippe Huget, Research on Multiagent System Design and Development

Introduction | Research challenges | Metamodelling | Domain-Specific Languages | Tools | Metamodel repository
Introduction | Research challenges | State of the art | YOSEMITE
Introduction | Research challenges | State of the art | JMS-based Middleware | P2P-based Middleware
Introduction | Research challenges | Interaction diagram | Agent diagram | Current work
Introduction | Getting started | Some examples | Associated tools
Introduction | Manufacturing | Timetable scheduling | ULSSIS

Research interests

Modelling multiagent systems

Modelling for every kind of artifacts is an important task since it will give shape to user requirements and discover technical flaws and user misunderstanding. Modelling in multiagent systems is still in its infancy and there is no common agreement around a notation as it is the case for classic software systems with UML. My work focuses on defining a notation that is enough exhaustive to cover the different kinds of cognitive agents but sufficiently flexible to allow designers to extend it for their purpose. My first attempt, part of the FIPA standardisation, was Agent UML. I am currently working on proposing a notation based on model-driven engineering easing the development and the extensibility. In this scenario, designers create specific metamodel and models and an editor is derived from this metamodel to allow designers (or users) to design their application.

Model-driven engineering for multiagent systems

Model-driven engineering is changing radically how software systems are developed. Software have to be released more frequently and for different kinds of platforms. Model is now the central element when developing software replacing code. Designers generate models for their applications then code is automatically generated for a specific platform. I adopt the same approach to develop multiagent systems and focusing on generating domain-specific languages for the development of multiagent systems. Currently, metamodels can be defined and an editor is generated for this specific metamodel. Model transformations are proposed to generate specific platform models then code is derived from these models. Main interests in this moment are to augment the quantity of code generated by considering the behavior part.

Multiagent systems for ULSSIS

Ultra-Large Scale Software Intensive Systems (ULSSIS for short) ( are software systems that could have several dimensions facing the scalability problem: number of users, number of software components, number of software interconnections, number of lines of code, etc. For these systems, multiagent systems could offer an elegant solution to some of ULSSIS challenges. I particularly study the impact and the advantages of using multiagent systems for the execution of ULSSIS.

Validation of multiagent systems (discontinued)

Validation of multiagent systems is still a subject underrepresented in multiagent system research, partly due to the absence of efficient tools that cope with agent features such as autonomy or asynchronous communication. I am mainly interested in considering the different alternatives available such as reachability analysis, model checking or testing and considering which one is the best. My first work was MABLE with Michael Wooldridge (

Rational decision making in autonomous agents (discontinued)

Autonomous agents need to make decision on what to do next based on the current state of the world. One of the main technique for this purpose is the BDI theory. In this area, I work on developing BDI agent models. Current work is focused on belief revision and its counterpart for multiagent systems called multi-agent belief revision.

Communication in multiagent systems (discontinued)

Communication is without doubt a first-class citizen in multiagent systems since it endorses coordination and simply information exchange. To this purpose, I defended an interaction protocol engineering in my Ph.D. thesis based on the communication protocol engineering. Latest work in the domain were the multi-party communication simulating human communications with speakers, hearers and eavesdroppers.

©2008 Marc-Philippe Huget