Requirements Model for Agent-Based Self-Configuration of Modular

Modular Assembly Systems

The formalisation of models that enable the clear and structured capture of the different aspects required for the configuration of MAS is the first knowledge contribution contained in the work. The justification for its existence is simple, without a clear model that can be computer interpretable it is not possible to establish any method to support the configuration process. Figure 3.4 provides an overview of the models required to enable the self-configuration methodology while highlighting the involved actors.

The actors identified are the equipment module supplier, the configuration expert and the system integrator. It is important to identify the actors since they are the source of all the information that is required to formalise the models. The analysis of the individual contributions in the MAS domain allows for the formalisation of these contributions. The proposed model will be established after an analysis of current state-of-the-art configuration procedures in the scope of the collaborative European project EUPASS.

In the proposed model seen in Figure 3.4, the equipment module suppliers will be required to supply their module description following a specific format that adheres to the common concepts and terminologies. Similarly the system integrator will also define the MAS requirements following a specific format that also adheres to the common concepts and terminologies. However, without the establishment of common concepts and terminologies it would not be possible to map the

46 requirements to the existing capability. Furthermore, the concepts will require updates as the domain evolves. To address these issues, it is proposed the creation of a new role of configuration expert that is able to add, change and update these concepts and terminologies. Finally a data model needs to be created for the solutions to be presented to the system integrator.

Equipment Modules XML Module Description XML Module Description Equipment Module Description Syst e m R e q u ire me n ts Self- Configuration Methodology Process Requirements Apply

Glue Place Cure

XML System Requirements Business Informtion System Constraints Feeder Skills: Interfaces: Description:

Feeder unit equipped with part location device.

Bay Port (Male) Part Port (Male) Feeding XML Module Description System Integrator Configuration Expert Adheres To Common Concepts and Terminologies Assembly Process Library Interface Library Adheres To Module Supplier

Module Supplier Module Supplier

Figure 3.4 - Overview of Requirements Model for Agent-Based Self-Configuration of Modular Assembly Systems

In sum, the requirements model for the self-configuration of MAS will include models for assembly process and interface libraries, for the definition of MAS requirements, for equipment module descriptions and for the description of MAS configuration solutions.

47

3.2.6 Agent Architecture for Distributed Self-Configuration

Methodology for Modular Assembly Systems

The creation of an agent architecture that is able to represent the aspects of the MAS configuration problem is the first step in the creation of the proposed bottom up distributed self-configuration methodology. The basic notion of this proposal is that agent technology can enable the creation of this methodology. For this, one needs to create agent types, roles and a structured hierarchy that is able to accurately structure the configuration problem.

The design of a multi agent architecture requires a structured approach. In the literature there are a couple of methodologies for the design of multi agent solutions. The majority of the existing methodologies are domain specific, however the GAIA methodology provides a good generic approach for the architecture design that has proven itself in computer science domain (Zambonelli et al. [123]). The design of the agent architecture based on the GAIA methodology requires firstly an analysis of the problem, namely the clear definition of the objectives and targets that the agent system has to address. The first step is the understanding of the requirements for such system, specifically the identification of what needs to happen and what information is required. The required information was already identified in the Requirements Model for Agent-Based Self-Configuration of Modular Assembly Systems. In this, requirements are established that have an impact on the objectives of modular assembly system configuration and reconfiguration. The main objective of the system is to provide valid solutions for the configuration and reconfiguration of MAS.

Once the objectives are defined, the next step in the design of a multi agent architecture is the analysis of the problem. This establishes the need for the definition of the agent types, roles and expected interactions.

The nature of the MAS paradigm provides a clear focus on functional decoupling of equipment module functionalities and standardised interfaces for interchange ability. This enables the formalisation of functional capabilities and connectivity constraints of the available modules hence allowing the mapping of requirements against available capabilities. This clearly identifies the two main agent types required for

48 the self-configuration methodology agent architecture based on the different objectives: the Equipment Module Agent and the Requirements Agent. This decoupling into two agent types, uses principles from blackboard architecture model, where two agent types come together to solve a problem, the difference being these will have a structure and common understanding of the relations between the two aspects of the configuration problem. These agents will provide the basic functionality required for solving the configuration problem. The Equipment

Module Agents provide representation of equipment modules, which can interact

with each other to establish collaborations that represent configurations. The

Requirements Agent will provide the objectives that motivate the Equipment

Module Agents interactions, while also being able to evaluate the solution based on the requirements established by the system integrator. Figure 3.5 provides a conceptual overview of the agent architecture for the self-configuration methodology, where all the agent types and respective hierarchies are established.

Figure 3.5 - Agent Architecture for Distributed Self-Configuration Methodology for MAS

Self-Configuration Methodology

Virtual Sandpit

Agent Environment Requirements Agent

Equipment Module Agent Performance Simulation Agent MAS Expert Agent

Equipment Module Description Syst e m R e q u ire m e n ts System Configuration

49 The nature of agent technology allows for the distribution of decision making processes that would be computer intensive through the creation of child agents, therefore taking advantage of distributed computing. The Performance Simulation

Agent is introduced into this architecture to provide a decoupling of one of the most

computer intensive problems, the simulation of given set of configurations for selections. It is easy to understand how the computer processing requirements grow exponentially if simulations for the performance characteristics are required for all solution possibilities.

The final aspect of the analysis of the configuration problem has to do with assessing the logical conditions of the configurations based on its internal knowledge model. The issue is, if this knowledge was built in to the configuration methods, e.g. the internal decision making models of Equipment Module Agents, future changes might require a complete change of the configuration methodology. Therefore, it is proposed that this knowledge is decoupled into the MAS Expert Agent, which can be changed or even replaced in order to cater for the evolution of this knowledge.