An agent-oriented methodology for hybrid intelligent system construction
- Publication Type:
- Issue Date:
In recent years, both researchers and practitioners have recognised the advantages of applying the agent-based paradigm to the development of hybrid intelligent systems (HIS). Yet, the number of deployed commercial agent-based hybrid intelligent applications is small. One of the reasons for this is the lack of practical methodologies for agent-based hybrid intelligent applications development. The aim of this thesis is to overcome this limitation. We have devised an agent-oriented methodology, called MAHIS (standing for Methodology for constructing Agent-based Hybrid Intelligent Systems). To avoid building MAHIS from scratch, we have followed the strategy of extending an existing well-known methodology in order to bridge the gap between the existing methodology and agent-based HIS construction. MAHIS has extended the capabilities of MAS-CommonKADS in agent-based HIS development. It is suitable for constructing agent-based HIS, as well as analysing and designing any open systems with hierarchical structure. Open system in this thesis means that the system allows for dynamic addition or removal of agents at run-time. MAHIS consists of eight models: Hybrid Strategy Identification Model, Organisation Model, Task Model, Agent Model, Expertise Model, Coordination Model, Reorganisation Model, and Design Model. These models are grouped into three levels: conceptualisation, analysis, and design. Both the Hybrid Strategy Identification Model and Reorganisation Model are newly developed models rather than from MAS-CommonKADS. At the same time, some existing models have been improved accordingly. The Reorganisation Model is the key model to support HIS and any open systems with hierarchical structure. It consists of category role, group roles, virtual organisation role, and dynamics rules. This model describes the hierarchical, dynamic, reusable, and unpredictable characteristics of HIS with virtual organisation, category, and group perspectives. Some previously developed agents can be reused by means of involving them in a new virtual organisation dynamically. The output of the Reorganisation Model is the specification of the dynamic platform which comprises middle agents and makes all agents and agent groups hierarchical and dynamic. A dynamic platform PAHIS (Platform for Agent-based Hybrid Intelligent Systems) has been developed. PAHIS not only verified the capability of MAHIS in dynamic platform construction, but also can be used as an infrastructure of agent-based HIS. It supports the dynamic addition and removal of group-based agents at run-time. A self-organising ring-based architectural model has been developed to organise middle agents in PAHIS. Moreover, the ring-based architectural model has been evaluated from the viewpoints of complexity, efficiency, extendibility, and availability. The ring-based architectural model is competent in agent-based systems with middle agents. The verification of MAHIS in HIS construction has been conducted by two successful case studies: "financial investment planning system" and "petroleum reservoir characterisation system". The former has verified MAHIS in constructing agent-based HIS with tight-coupling hybrid strategy. The latter has verified MAHIS in constructing agent-based HIS with loose-coupling hybrid strategy. PAHIS has been employed in these two systems. The evaluation of MAHIS with the enriched evaluation framework which was originally proposed by Cemuzzi and Rossi has been completed by comparing it with Gaia and MAS-CommonKADS. The reorganisation attributes have been added into the evaluation framework. The evaluation results have indicated that MAHIS is preferable over Gaia and MAS-CommonKADS, especially in the construction of agent-based HIS.
Please use this identifier to cite or link to this item: