在早期的建築設計階段，設計師的想法被看作是一個黑盒子。設計師從他們心中隱喻的想法來構成他們的設計。但是這些設計成果，特別是概念成型階段的期間，並不能全然地反映出設計師的行為。實際上，這些情況發生在設計事務所、現場施工和設計課程教育中，如同設計的反應、設計的次序或是設計的事件在不同的設計概念裡出現。
設計領域可以隨著這個設計的象徵－想像力實現的程序來擴大和顯示。動態的角色扮演以動態想法地圖為基礎，簡化成為多重角色扮演程序。設計議題或設計概念和其他的議題或概念在特定的目標下進行競爭，設計概念可以由具優勢的設計想法再進行演化。應用兩種學習型式可以實現它們：競爭學習和演化學習。這兩種學習在設計的序列衍生提取了不同的角色以及在內的設計知識。根據這些概念以及運算上的支持，本研究建立一個能進行競爭和演化學習的競技場系統－Competing and Evolving Learning Arena (CELA) system。以模擬角色扮演在分散式代理人的基礎下的設計行為。一個特別的架構稱作"仲裁者"，在競爭學習和演化學習方面能讓仲裁結果更為清楚。仲裁者也處理以及操作系統整體的模擬。各個系統內部的過程表現在特定的方法，機制和運算架構。

In the early phase of architecture design and the process stage, the mind of designers is treated as a black box. Designers are constructed from their mental metaphors and the outcomes do not follow their conducts, performances and pedagogy closely, especially during concept developing stage. In reality, these situations do happen in the design studio, and on scene construction, and design class education that are like the reaction sequences or events occurred under different design concepts.
Design domain can be widened with the design metaphor－imaginative reality process is manifested. Dynamic Role Interplay is simplified as multi-roles play process which based on the Dynamic Idea Map. The design issues that are either concepts competing with others for a particular objective or the design concepts can be evolved because of superiority ideas. Two types of learning can be applied to realize them: competing learning and evolving learning. They abstract the different roles in sequence generations of design, and then the knowledge within. According to these concepts and computational supports, a Competing and Evolving Learning Arena (CELA) system with a simulation on distributed agent-based role-interplay environment is established. A special framework called arbitrator to make the arbitrations clarified, in the aspect of competing learning and evolving learning. The arbitrator also manipulates the whole simulation. Each process behaves within the metaphors of specific approaches, mechanisms and computation framework.

LIST OF FIGURES
CHAPTER 1
INTRODUCTION
1.1 MOTIVATION
1.2 DESIGN PROBLEM STATEMENT
1.2.1. Why "imaginative reality" is considered
1.2.2. How “imaginative reality” is discussed
1.2.3. Learning processes confronts with the design problems or software integrations
1.3 OBJECTIVES
1.4 SCOPE
1.5 APPROACH
CHAPTER 2 RELATED RESEARCH
2.1. DISTRIBUTED LINKING IDEAS
2.1.1. Acting Role Model (ARM)
2.1.2. Dynamic Idea Maps (DIM)
2.2. COMPUTATIONAL SYSTEM FRAMEWORK AND AGENT TECHNOLOGY SUPPORT
2.2.1. Integrated Building Design Environment system (IDBE)
2.2.2. Design Objects (DOs)
2.2.3. Dynamic Agent Role Interplay System (DARIS)
2.3. LEARNING, NATURAL SELECTION OPERATORS AND NETWORK STRUCTURE
2.3.1. Agent learning theories
2.3.2. Genetic competing concepts and Evolutionary system
2.3.3. Agent network structures
2.4. SUMMARY
CHAPTER 3 ANALYSIS
3.1. REIFICATION METHODS AND STEPS
3.2. DISTRIBUTED LINKING IDEAS
3.3. COMPUTATIONAL SYSTEM FRAMEWORK AND AGENT TECHNOLOGY SUPPORT
3.4. COMPETING LEARNING STRATEGIES AND EVOLVING LEARNING STRATEGIES
CHAPTER 4 THE FRAMEWORK
4.1. CONCEPT OF THREE LAYERS
4.2. LAYERS OF ADAPTIVE CHARACTERISTICS
4.2.1. Representation layer
4.2.2. Communication layer
4.2.3. Learning layer
4.3. CONCEPT OF ARBITRATOR
4.4. PREFERENCES TRAINING AND SEARCH
4.4.1. Preferences training for search
4.4.2. Search for preferences training
4.5. THREE "SELECTIONS"
4.5.1. Three selections in Arbitrator
4.5.2. Three selections in Human
4.5.3. Three selections procedures
4.6. DATA MODEL
4.6.1. ICF Maps
4.6.2. Dynamic ideas linking storage
4.7. COMPETING LEARNING MODEL
4.7.1. Concept of Arbitrator in Competing
4.7.2. Competing learning events from simulation
4.7.3. Internal interplay
4.7.4. External interplay
4.7.5. Genetic competing method
4.8. EVOLVING LEARNING MODEL
4.8.1. Concept of Arbitrator in Evolving
4.8.2. Internal interplay
4.8.3. External interplay
4.8.4. Evolutionary computational strategies and methods
4.8.5. Evolutionary computational procedures
CHAPTER 5 THE IMPLEMENTATION
5.1. IMPLEMENTATION DIFFICULTIES
5.2. TECHNICAL CONSIDERATION
5.2.1. Distributed agent environment and Agent Communication
5.2.2. Idea Entities and Knowledge Representation
5.3. CELA SYSTEM
5.3.1. Implementation layer
5.3.2. System Architecture
5.3.3. Agents Purposes and the Classes Diagrams in Interplay Process
5.3.4. Implementation Selections with Arbitrator
5.3.5. Implementation Search and the Information Model
5.3.6. Implementation Distributed Design Ideas Representation
CHAPTER 6 SIMULATION AND SCENARIOS
6.1. SIMULATION ANALYSIS
6.2. DESIGN EDUCATION
6.2.1. Design Case
6.2.2. Initialization the Play
6.3. DESIGN STUDIO
6.3.1. Design Case
6.3.2. Internal Competing / Evolving Learning and the Sub-Scene Event
6.4. DESIGN CONSTRUCTING
6.4.1. Design Case
6.4.2. External Competing / Evolving learning
CHAPTER 7 CONCLUSION AND FUTURE WORKS
7.1. DISCUSSION AND SUMMARY
7.2. CONTRIBUTIONS
7.3. FUTURE WORKS
REFERENCES
APPENDIX
APPENDIX A 22ND CIB -W78 2005 PUBLISHED PAPERAPPENDIX B EISTA 2005 PUBLISHED PAPER
APPENDIX C CAADRIA 2005 PUBLISHED PAPER

Aly, S. and R. Krishnamurti (2002). Can doors and windows between design team players? Agents in Design 2002, Key Centre of Design Computing and Cognition, University of Sydney, Agents in Design 2002.
Baeck, T., D. B. Fogel, et al. (2000). Evolutionary Computation 1: Basic Algorithms and Operators (Evolutionary Computation), Institute of Physics Publishing.
Banzhaf, W., P. Nordin, et al. (1997). Genetic Programming: An Introduction: On the Automatic Evolution of Computer Programs and Its Applications, Morgan Kaufmann.
Bowling, M. and M. Veloso (2002). "Multiagent learning using a variable learning rate." Artificial Intelligence 136(2): 215-250.
Chang, T.-W. (2002). Pedagogy or Andragogy: a role-interplaying approach for digital media learning. 36th Annual Conference of the Australian and New Zealand Architectural Science Association (ANZAScA) 2002 Conference: Modern Practice of Architectural Science: from pedagogy to andragogy/, Geelong, Victoria, Australia., School of Architecture and Building, Deakin University.
Chang, T.-W. (2004). Dynamic Role Interplay in Design Process, Graduate Institute of Architecture, National Chiao Tung Univesity: (in preparation).
Chang, T.-W. (2004). Modeling Situated Generation Using Role-Interplay: From situated interplay to generative interplay. Generation CAD Systems Symposium (G-CADS 2004). Carnegie Mellon University, School of Architecture, G-CADS.
Chang, T.-W. and I.-C. Lai (2004). Dynamic Agent-Based Role Interplay System. Taipei, National Science Council: technical report.
Chang, T.-W. and I.-C. Lai (2005). Implementing Distributed Interaction with Dynamic Agent Role Interplay System (DARIS). Automation in Construction: submitted.
Clausen, C. and H. Wechsler (2000). "Quad-Q-Learning." IEEE Transactions on Neural Networks 11(2): 279-294.
Coley, D. A. (1997). An Introduction to Genetic Algorithms for Scientists and Engineers, World Scientific Publishing Company.
Eiben, A. E., J. E. Smith, et al. (2003). Introduction to Evolutionary Computing, Springer.
Fenves, S., U. Flemming, et al. (1994). Concurrent Computer-Integrated Building Design, Prentice Hall.
Friedman-Hill, E. (2003). JESS in Action: Rule-Based System in Java, Manning Publications.
Gero, J. S. (2000). Computational models of innovative and creative design processes. Technological forecasting and social change, Elsevier Science.
Goldberg, D. E. (1989). Genetic Algorithms in Search, Optimization, and Machine Learning, Addison-Wesley Professional.
Holland, J. H. (1992). Genetic algorithms. Scientific American. 1: 66-72.
IEEE Computer Society Standards Organization. (2005). "Foundation for Intelligent Physical Agents (FIPA)." from http://www.fipa.org/.
Kao, K.-T. and T.-W. Chang (2005). Competing Learning and Evolution in Design Education Support. Orlando, Florida, USA, The 3rd International Conference on Education and Information System: Technologies and Applications: submitted.
Kao, K.-T., T.-W. Chang, et al. (2005). Learning Construction Decision Making with Arbitrator - Competing and Evolving in Dynamic Role Interplay. Technische Universität Dresden, Germany, 22nd CIB -W78 conference Information Techonology in Construction: submitted.
Katagiri, H., K. Hirasawa, et al. (2000). Genetic network programming - application to intelligent agents. Systems, Man, and Cybernetics, 2000 IEEE International Conference, Nashvile, IEEE International Conference.
Lai, I.-C. (2004). Dynamic Linkages between Ideas and Cases. Ruwenberg, Developments in Design and Decision Support Systems in Architecture and Urban Planning: 225-238.
Lai, I.-C. and T.-W. Chang (2004). Dynamic Interactions Between Users and Information in a Co-existence Space. The eCAADe 2004 22nd conference, Copenhagen, Denmark.
Lai, I.-C., K.-T. Kao, et al. (2005). Role Playing for Linking Ideas in the Idea Association Process. Proceedings of 10th Conference on Computer Aided Architectural Design Research (CAADRIA), New Delhi (India), CAADRIA.
Mitchell, M. (1998). An Introduction to Genetic Algorithms. Cambridge, The MIT Press.
Mitchell, T. M. (1997). Machine Learning, McGraw-Hill Science/Engineering/Math.
Mulky, A. R. and F. Ozel (2004). An Evolutionary System to Generate Multi-Objective Architectural Solutions. Generative CAD Systems Symposium 2004 (G-CADS 2004). Carnegie Mellon University, G-CADS 2004.
Oxman, R. (1994). Precedents in design: A computational model for the organization of precedent knowledge. Butterworth Heinemann, Design Studies. 15: 141-157.
Sutton, R. S. and A. G. Barto (1998). Reinforcement Learning: An Introduction (Adaptive Computation and Machine Learning), The MIT Press.
Tilab. (2005). "JADE (Java Agent Development Framework)." from http://jade.tilab.com/.
Weiss, G. (2000). Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence, The MIT Press.
Yardley-Matwiejczuk, K. M. (1997). Role Play: Theory and Practice, SAGE Publications.