Software systems characteristics and expectations have fundamentally changed in the past decade. Increasing both in size and complexity, actual software systems are expected to be distributed, open and highly dynamic. Multiagent systems are emerging as probably one of the most adapted software engineering paradigm for developing complex software systems [43, 78]. However the current practice of Multi-Agent System (MAS) design tends to be limited to individual agents and small face-to-face groups of agents that operate in closed systems . This practice seems in contradiction with the current evolution of software systems requirements and what previously reported about complex systems.
According to Simon , complex systems often (if not always) exhibit a hierarchical configuration1. And the idea that the architecture of a complex systems can be explained and understood using a “hierarchical organization structures” is shared by a relevant number of scientists [25, 48, 70, 76]. Several metamodels and methodologies have been proposed for MAS [6, 40]. However, most of them see agents as atomic entities. Considering agent as composed entities and thus enabling a modeling of nested hierarchies (and their associated dynamics) may offer a more adapted and manageable way to model complex systems.
Giving this landscape, we advocate the use of holonic multiagent systems (HMASs) combined with an organisational approach for developing complex software systems. This paper introduces an agent-oriented software process for engineering complex systems called ASPECS. The process can be considered an evolution of the PASSI  methodology for modelling HMAS systems and it also collects experiences about holons design coming from the RIO approach . The construction of the new process has been performed according to the situational method engineering paradigm [8, 38, 64] and the approach described in . The description of the method adopted for building the ASPECS process is out of the scope of this paper and the only element that can be considered relevant for this context is that the MAS metamodel adopted by the new process has been the first adopted choice. For this reason the description of each phase of ASPECS will start from the portion of MAS metamodel instantiated/refined by its activities.
ASPECS is based on a holonic organisational metamodel and provides a step-by-step guide from requirements to code allowing the modelling of a system at different levels of details using a suite of refinement methods. Using a holonic perspective, the designer can model a system with entities of different granularities. He can recursively model subcomponents of a bigger system until he achieves a stage where the requested tasks are manageable by atomic easy-to-implement entities. The concept of holon is central to the proposed discussion and therefore a definition of what is a holon could be helpful. In multiagent systems, the vision of holons is someway closer to the one that MAS researchers have of Recursive or Composed agents. A holon constitutes a way to gather local and global, individual and collective points of view. A holon is a self-similar structure composed of holons as sub-structures and the hierarchical structure composed of holons is called a holarchy. A holon can be seen, depending on the level of observation, either as an autonomous “atomic” entity or as an organisation of holons (this is often called the Janus effect ). Holonic Systems have been applied to a wide range of applications like Manufacturing systems [54, 77], Health organisations , Transportation , Adaptive Mesh Problem , Cooperative work  just to mention a few of them. Thus it is not surprising that a number of models and frameworks have been proposed for these systems, for instance PROSA , MetaMorph [54, 68]. However, most of them are strongly attached to their domain of application and use specific agent architectures.
For a successful application and deployment of MAS, methodologies are essential . Number of methodologies and metamodels with a clear organisational vision have been already proposed: metamodels like AGR , RIO , MOCA , and methodologies like GAIA , INGENIAS , ISLANDER , MESSAGE , or SODA . Most of these methodologies recognize that the process of building MASs is radically different from the process of building more traditional software systems. In particular, they all recognize (to varying extents) the idea that a MAS can be conceived in terms of an organized society of individuals in which each agent plays specific roles and interacts with other agents [42, 78]. As pointed out by Ferber [26, 27], organisational approach offers a number of advantages and can contribute to agent-oriented software development in the following points: heterogeneity of languages, modularity, multiple possible architectures, security of applications. The objective of the proposed work consists in trying to gather the advantages of an organisational approach and those of the holonic vision to model complex systems; the result will be a set of organisationoriented abstractions that have been integrated into a complete methodological process called ASPECS.