Article Preview
Top1. Introduction
New trends have been proposed in the design and the development of complex systems in terms of modelling/verification approaches, languages and tools. The choice of the most rigour expressive language and the powerful support tool for a given problem is still a challenge for designers. Formal methods and model-driven approach (MDA) are both proposed as adequate combined solutions to design and implement correctly appropriate designer’s tasks. The necessity for rapid evaluation of a given system during its development phase is also a promising and enriching approach, which allows dynamically rectifying and refining the projected product and its observation at various points of its development.
A multi-agent system (MAS) is considered as a natural solution to overcome the system’s complexities. The general principle of MAS consists of the relevant division of the overall system into subsystems in order to divide the overall complexity of the considered system.Many researchers investigated the context of the formalisation of multi-agent systems. The chosen formalism should present mechanisms to express the key characteristics of MAS. In this way, the Agent Petri Nets formalism (APN) (Marzougui et al., 2010) is an extension of the Petri Nets dedicates to express several aspects and characteristics of MAS in a formal way.
The APN provides a set of functions to model agent’s behaviours and interactions. In addition, the graphic representation of APN facilities the comprehension of the considered model.
Recently, the model-driven approach paves the way for the designer. MDA is largely used to facilitate, accelerate the development process of applications, and to increase the efficiency of the modelling activities (Atkinson & Kuhne 2003). Tools and approaches are proposed to promote the definition of new languages of specific domains using the meta-language concepts. Furthermore, big facilities are provided by the MDA to build the adequate software of a given problem and facilitate designers' tasks at the same time.
To experiment an Agent Petri Nets model, in reality, convenient tools that allow editing and simulating this model are elementary requirements. Providing simulation environment can help people to experiment quickly their models. The implementation of model simulation is easier than using the proof technique to check the model (Prat 2017). As an evolution of the implementation of graphical modelling frameworks, CINCO (Naujokat et al., 2018) provides a holistic approach that facilitates the generation of specific graphical tools and promotes the integration of simulation mechanisms; simplicity and specialisation are two crucial characteristics of the CINCO tools.
The aim, in this research, is to propose an approach that allows the manipulation of Agent Petri Nets models using the new trends in the area of development of graphical tools. Specifying and simulating APN model are both considered in this approach. To achieve this goal, first, the capabilities of a full generation of CINCO products to develop a specific tool for APN are exploited. Secondly, an approach to execute Agent Petri Nets models is proposed. Users will experiment with their APN models in the runtime environment. Java is the target programing language used to implement the simulation code. To facilitate and accelerate the development and the evaluation of system models, the Agent Petri Nets (APN) model is proposed. In the proposed work, the CINCO tooling suite is used to full generation of a specific graphical modelling tool for the APN models. So, the simulation code is added to the generated code. Therefore, with help of this tool, the user can build models based on Agent Petri Nets formalism and at the same time execute these models. To illustrate and explain the proposed approach, an example of a traffic light adaptive system is considered and discussed in this paper.
The rest of the paper is organized as follows: Section 2 presents some backgrounds and related works on modelling and verification tools for domain-specific languages. Section 3 gives a brief overview of the Agent Petri Nets formalism as a specific language for modelling multi-agent systems, while Section 4 describes the key concepts in the CINCO product. Section 5 is devoted to present the proposed approach. In section 6, an application example is presented. The last section concludes this paper with some perspectives for future researches.