*3.4. MAS Design and Architectures*

3.4.1. MAS Design Methodologies

According to [109], the MAS design usually consists of


Although the majority of the selected literature not specifically present their phases of the MAS design methodology, the introduction of the MAS architecture/structure in their cases is more or less according to the Gaia methodology (shown in Figure 5). The Gaia methodology is popularly adopted for the analysis and design of the agent-based system, it is used in [23,24,50]. Some other similar methodologies are also used for the agent-based system design, e.g., High-Level and Intermediate Models for Agent-oriented Methodology (HLIM), Modelling Agents and their environment (AUML), MASE [24].

Another MAS design methodology proposed by the IEEE PES MAS working group(http://sites. ieee.org/pes-mas/agent-technology/design/) is mentioned in [17]. This MAS design methodology is proposed by [110] with six stages, and each stage of the methodology produces material that is input to the next stage (shown in Figure 6):

• Requirements and knowledge capture stage: the MAS design usually begins with a particular problem. To solve this problem, this stage specifies the system requirements and capture the knowledge needed to fulfill those requirements. The system requirements and captured knowledge is the input to the next stage.


**Figure 5.** The conceptual framework of the Gaia methodology [111].

**Figure 6.** Agent design methodology stages and their output used during the design of the PEDA (Protection Engineering Diagnostic Agents) system [110].

Some MAS design also defines layers of the MAS architecture, e.g., [23] that the MAS system architecture includes two layers: the management layer and the subjacent execution layer. The management layer is responsible for functionalities that can be considered general in the frame

and are covered by the agent control systems. The subjacent execution layer employs the automation agents' tasks.
