AN AGENT-BASED CONFLICT MANAGEMENT SYSTEM FOR CO-OPERATIVE DESIGN ENVIRONMENT

Transcription

1 AN AGENT-BASED CONFLICT MANAGEMENT SYSTEM FOR CO-OPERATIVE DESIGN ENVIRONMENT A Gayretli BSc MSc PhD and S Kucukgokoglan BSc MPhil School of Mech., Mat., Manu. Eng. and Management, The University of Nottingham, University Park, Nottingham, NG7 2RD, UK ahmet.gayretli@nottingham.ac.uk Abstract Increasing complexity of today's customer products requires co-operative design process (CDP). The success of this process needs the effective resolution of conflicts between different design domains during the design process. Conflict management is still an important issue in CDP and needs effective tools and methodologies to avoid from iterations adding cost. This paper describes an agent-based conflict resolution system, which has been linked to a co-operative design environment. It is based on constraint network programming in order to help the designers to efficiently deal with conflicts during the design process. The developed system has been tested on an automotive component. Results showed that it could provide effective monitoring, detection and resolving of conflicts leading to costly iterations in the early design stages. Keywords: Co-operative Design, Conflict Resolution, Monitoring, Networks, -Based Systems 1. Introduction Co-operative product development has become necessity for the design of complex products. In order to successfully implement it conflicts between different design areas have to be detected and resolved during the design process. Conflict management is still an important issue in the co-operative design process and needs effective tools and methodologies (Gayretli and Abdalla, 1999). Since co-operative product development requires the involvement of a number of engineers from different departments, it is necessary to critically consider several tasks; overall co-ordination, control, consistency, and data integrity, negotiation and collaborative decision making (Prasad et al. 1993). The acquisition of new knowledge and sharing information between separate groups of designers are the key factors essential for co-operative product design and it is a very beneficial strategy for achieving advanced product design solutions (Yoshimura and Yoshikawa, 1998). However, conflicts may often exist between members of the design team separated one from another by a number of obstacles, which include technical disciplines, culture, distance and lacks of skills for conflict resolution and understanding of the co-operative design process. Since almost 80 per cent of the product cost are determined at the design stage conflict resolution plays an important role in reducing lead time and achieving successful product designs. Many tools and methodologies have been developed for supporting conflict management. An agent based system have been developed by based on the distribution of the preferences and constraints from a supervisor agent to sub-agents, who utilise their local expertise within the defined global context (Balasubramanyam and Norrie, 1996; D Ambrosia et al. 1996). However, any changes in a design variable require extensive communication between agents, hence longer time to reach the optimum solution. 1

2 networks have been implemented to help the designer to improve product designs by avoiding conflicts. The constraint network has a number of advantages: 1. flexibility to allow the design problem to be analysed from different points of view. 2. providing designers with the design of products with incomplete information. 3. ability to handle large variety of life-cycle information. However, complex designs cannot easily be represented in terms of constraints and it is also very time-consuming. networks have been implemented for resolving conflicts by several authors (Bowen, 1997; Hayes and Sun, 1994). Pena-Mora et al. (1994) presented a framework called SHARED-DRIMS for resolving conflicts within an organisation where many different types of professionals, who must interact and communicate with one another. The proposed approach includes mechanisms for checking interactions and prompting hypotheses about the reasons for the interaction. Negotiation has been proposed for an effective technique for conflict resolution (Harrington et al. 1995). The negotiation strategy in general includes five steps: identification of conflict, search for cause of the conflict, analysing conflict situation, selection of conflict resolution strategy, applying conflict resolution strategy. Since, no particular strategy can be applicable to all conflict situations conflict resolution requires a good understanding of the problem and a suitable selection of a resolution strategy to eliminate it. Blackboard systems also have been used for dealing with conflicts. For example, Susan et al. (1996) presented blackboard architecture for supporting the integration of heterogeneous collaborative agents. The blackboard architecture comprised of three major elements: the blackboard database, control mechanism and a set of experts. In this paper a new agent-based conflict resolution approach using constraint network is presented in order help the designers to co-operatively design complex components by effectively dealing conflicts during the design process. In Section 2 the developed prototype conflict resolution system, being linked to a co-operative design environment is briefly described in order to assist the design team in avoiding conflicts, conflict detection, monitoring and resolution. The paper concludes with the analysis of an automotive component showing the capability of the system. 2. The Co-operative Design Environment Figure 1 illustrates the generic co-operative design environment for the design of complex products. It enables designers to consider the issues of the entire product life cycle concurrently in the early design stages. It is made up of a number of elements, which interact with one another. Designers carry out various analyses such as manufacturing capacity checking, process selection and time/cost estimation. He/she has to interact with a CAD system in order to obtain the topological and geometrical attributes of the component so that the analysis process can begin. Design tasks are distributed to individual agents, who have the expert knowledge to accomplish their individual tasks such as conflict resolution. The consistency management system deals with conflicts amongst the agents. It has the ability to monitor, detect and resolve inconsistencies by proposing solutions to designers. The constraint-based system provides constraint propagation through the constraint network representing collections of constraints. If a design value violates any constraint the conflict resolution system will detect this violation, and show its reason to the designer, and finally give suggestions for resolving the conflict. The environment has a optimisation module, which includes a rule-based algorithm and process feature table to select feasible processes for the component subject to the predefined constraints such as tolerances, availability, time and cost. 2

3 Market Requirements & Design Specification CAD System Pro/Engineer Product Data Designer User Interface DFM Menu Design Panel Type scri pt Window Intelligent -Based System Design Tasks PDS Manufacturing Capacity Material Cutting Tool Machine Tool Prototype Te sting -based System User Input M. Capacity Material S. S. Machine P rototype Ti m e /C ost Esti m ation Optimisation Rule- based Feature Algorithm Table KBS Knowledge Representation Methods Feature-based rep. Frames KEE P roduction OOP rules THE SYSTEM PROVIDES * Reduced product cost and development time by avoiding costly design conflicts * Maximum and economical use of of available manufacturing facilities * Feasible process selection and optimisation * Full analyses of of various design tasks such as DFM, process selection and time/cost estimation * Consideration of various life cycle issues in in the early design stages Inference Engine Forward Chaining Backward Chaining OOP Knowledge Source s Experts Machining Management Monitoring Conflict Detection Explanation System Conflict Resolution -based System Variables & propagation satisfaction violation Knowledge Acquisition Material Time/Cost Tooling Estimation Figure 1 The Co-operative Design Environment 2.1. An -Based Conflict Resolution Approach Using Contraints Network Different design tasks, which include material selection, manufacturability analysis, process selection and optimisation, need a huge amount of information to be accessed and shared in the knowledge base. This requires that the information transferred from one area to another is consistent. The addition of new information is also required to the knowledge base, in order to accomplish specific tasks. As shown in Figure 2, several agents represent various design analyses. Each agent has a task to carry out and requires a common knowledge base to access the necessary information. An agent is an entity capable of solving locally generated problems through communication with other agents. s have responsibilities for solving a given task in a given problem, such as process selection and/or capacity checking. They include a limited amount of program for dealing with a given sub-task, so that the given task can be executed faster and at less cost. Each agent should interact with one another and exchange information between other agents, in order to accomplish its own task. In order to ensure consistency in the constraint network, any new information from the designer or agents is propagated by the constraint-based system, which checks whether or not the new information causes constraint violation. Figure 3 3

4 shows how agents share information, while consistent information flow is achieved. If the design input violates any constraints of the agent, it will be detected by the consistency agent. A message will be then sent to a method (of a slot) carrying a small program that is a function of LISP and is responsible for conflict resolution. Alternative solutions will be provided to the designer, by a menu, from which a design agent has to make selections. He/she may be asked to write the necessary information at the user-prompt. The reasons for conflicts in detail will be shown by the explanation system. Design Feature Manufacturing Capacity CONSTRAINT-BASED Manufacturability Time/Cost Estimation SYSTEM Knowledge Base Rules CONSTRAINT-BASED SYSTEM es Cutting Tool Machine Figure 2 The -Based Conflict Resolution Approach Monitoring Violations Explanation System Conflict Resolution Design INPUT Rules Tasks _1 OUTPUT _2 Figure 3 Information Flow between s Conflicts and : represent certain limitations or restrictions on the design variables. For instance, from a process planner's point of view, the tolerance of a hole could be a constraint and the product cost another. In this research, design and manufacturing requirements are formulated as a set of constraints in slots of a unit, and in the production rules. Conflict is defined as "disagreement between two or more view points on some decision or values proposed in a design" (Pruitt, 1981). Our definition in this paper is "violation of constraints imposed on variables by values". have been used for removing inconsistent values from the domains of variables. 4

5 Management: Figure 5 shows how the conflict resolution system manages the decision-making process, deals with conflict situations and provides justification of decisions made on designs. It includes a mechanism for the monitoring and detection of conflicts and gives warnings and explanations to the designer. Any value for a variable is propagated through the contraints network. If a constraint violation exist a suitable strategy for resolving conflicts will be applied to ensure design consistency in the constraint network and design output. The designer is provided with the ability to take necessary actions to resolve conflicts and monitor design violations on the design consistency panel. The flowchart for conflict resolution is shown in Figure 4. Tooling Tooling Network Capacity Capacity Input Propagation Propagation Monitoring Monitoring Other Other Domain Domain Satisfaction Satisfaction Violation Violation Conflict Detection Explanation System Design Design Panel Panel Time/Cost Time/Cost Estimation Estimation Conflict Resolution Suggestions Suggestions to to the the User User Optimisation Optimisation Figure 4 The Flowchart for Conflict Resolution 3. Example The co-operative design environment is shown in Figure 6 and evaluated through the design of a cylinder head (Figure 5). An example of how it could be used to provide feedback to designers about various design tasks and conflicts are presented. During the analyses conflicts were arisen from different design domains due to the constraint violation. However, they were effectively monitored, detected and resolved as shown in Figure 8 and 9. T_Hole_2 Slot_2 Figure 5 A 3-D Solid Model 5

6 Figure 6 The Co-operative Design Environment Conflict Resolution through Intervals and Cardinality: The intervals were used effectively for representing and reasoning on the design parameters. Interval values were immediately propagated to detect potential violations before design analyses commenced. This provided the designer with the capability of avoiding possible conflicts very early in the design stage. Another attribute of slots was cardinality, which permitted specifying the number of values of slots. Figure 7 shows how the cardinality could be used to restrict the number of values of the set-up time of an EDM process. The designer was informed of the cardinality violation and not allowed the addition of a second value to the slot set-up in the EDM process. This reduced conflicts, which were likely to arise at the later stages of the design process. Figure 7 Conflict Resolution in The Early Design Stages Monitoring: Monitoring constraint violations helped him to understand what was really happening during the decision-making. The consistency in the constraint network was monitored via the design consistency panel and in the Typescript Window. The panel consisted of a number of active images linked to the slots, which included the value of a design variable (in this case, the production size), which was monitored (Figure 6). When constraints such as the maximum manufacturing capacity and the maximum length were violated, this was shown in the consistency panel and in the Typescript Windows (Figure 6 and 8). The reasons for the conflicts were shown in the form of text. The designer was provided with the ability to solve the inconsistencies through the consistency agent during the analyses. Any changes made on values were immediately highlighted on the consistency panel (Figure 6). This provided him with the ability to monitor inconsistencies. No violations were monitored when all the constraints were satisfied. 6

7 Figure 8 Conflict Resolution for the Component's Length Conflict Detection and Resolution: When the consistency agent detected inconsistencies or when any constraints were violated such as the one, which existed between the predefined requirement and production capacity, the reasons for the inconsistencies were immediately generated in the Typescript Windows. The consistency agent then generated suggestions for enabling the user to resolve the conflict. Figure 8 shows how the developed system detected a conflict between the length of the component (100 mm) and the manufacturing capacity (10 mm > length <100 mm). This conflict was also shown as "OUT OF RANGE" on the design consistency panel (Figure 6). The designer was suggested to change the length into a value between 10 mm and 100 mm. Then, the programme stopped running for a while and waited for information to be typed by the user at the prompt window. He/she modified the length into 99 mm. The new information was propagated again. Since, neither violations in the design consistency panel nor warnings were reported the manufacturing capacity checking was complete. Finally, the results of the manufacturing capacity checking were displayed in the Typescript Windows for the user's information. Also, this modification and the new length were immediately reflected on the design consistency panel as "WITHIN THE RANGE" (Figure 6). Communication: Communication was one of the most important issues in the cooperative design environment. There were interdependencies between design activities represented by the agents. This required interactions and information sharing in order to effectively carry out design analyses. To provide efficient information processing, amongst agents in order to achieve their individual goals without any conflicts, the agents shared an object-oriented knowledge base, which was easily accessible to each agent. Design analyses were sequentially carried out. However, the consistency agent was active during the analyses, in order to efficiently deal with constraints violation. It provided the designer with the capability to detect, display and resolve conflicts and work co-operatively with other agent in order to manage interdependencies between design domains, so as to achieve a successful design that satisfied all the constraints. 4. Conclusions An agent-based conflict resolution system has been developed and linked to the cooperative design environment in order to carry out the design tasks, by avoiding design conflicts between the different design domains. The system provides the design team with an effective management of information exchange and decision making within agents (design areas). This is based on the agent-based approach using constraint networks. This system enables the designer to consider various critical tasks (overall co-ordination, control, consistency, and data integrity) without costly-design iterations caused by conflicts. Design inconsistencies between different design domains have been overcome by 7

8 the conflict resolution system, which in turn reduces product development time. The designers could monitor inconsistencies through the consistency panel linked to the userinterface. It is comprised of three major elements; concurrent engineering menu and conflict resolution, design consistency control panel for monitoring inconsistencies and typescript window. The agent-based design consistency approach ensures that information transferred for one area to another is consistent. in the constraint network has been achieved by the consistency agent responsible for dealing with conflicts. The designers are provided with the ability to monitor inconsistencies visually via the user interface. References Balasubramanian, S. and Norrie, D., H., (1996) " A Multi-agent Architecture for Concurrent Design, Planning, Routing, and Scheduling", Concurrent Engineering: Research and Application, Vol.4, No.1, pp Bowen, J, (1997) Using dependency records to generate design co-ordination advice in a constraint-based approach to concurrent engineering, Computers in Industry, Vol. 33, pp D Ambrosio, J, Darr, T, and Birmingham, W., Hierarchical Concurrent Engineering in a Multiagent Framework Concurrent Engineering: Research and Applications, Vol.4, No.1, pp , Gayretli, A and Abdalla, H, (1999) " A Prototype -Based System for The Automation and Optimisation of Machining es", Journal of Manufacture-Part B, Vol. 213, pp Harrington, J.V., Soltan, H., and Forskitt, M., Negotiation in a Knowledge-Based Concurrent Engineering Design Environment, Expert Systems, Vol. 12, pp (1995). Hayes, C C and Su, C S, (1995) Using a Manufacturing Network to Identify Cost-critical Areas of Designs Artificial Intelligence for Engineering Design, Analysis and Manufacturing, Vol.9, pp Pena-Mora, F, Sriram, R D, and Logcher, R., Conflict Mitigation System for Collaborative Engineering Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 9, , Prasad, B, Morenc, R S and Rangan, R M, (1993) " Information Management for Concurrent Engineering: Research Issues" Concurrent Engineering: Research and Applications, Vol.1, No. 1, pp Lander, S E and Corkill, D D, (1996) "Designing Integrated Engineering Environments: Blackboard Based Integration of Design and Analysis Tools" Concurrent Engineering: Research and Applications, Vol. 4, No. 1, pp Yoshimura, M and Yoshikawa, K, (1998) Synergy Effects of Sharing Knowledge during Co-operative Product Design Concurrent Engineering: Research and Applications, Vol. 6, No. 1, pp Pruitt, D G, (1981), Negotiation Behaviour, Academic Press. 8