Links to document based files:

• Entire Document (195 pages, 1124 KBytes)
• Document in pieces
  • Abstract, Dedication, Acknowledgements, Table of Contents, List of Tables, List of Figures, List of Listings (9 pages, 95 KBytes)
  • Chapter 1: Introduction (9 pages, 36 KBytes)
  • Chapter 2: Decision-Making Software (30 pages, 178 KBytes)
  • Chapter 3: Software Agents (20 pages, 125 KBytes)
  • Chapter 4: State-Based Communication Mechanism (25 pages, 179 KBytes)
  • Chapter 5: Component-Based Decision-Making (41 pages, 209 KBytes)
  • Chapter 6: Analysis (14 pages, 69 KBytes)
  • Chapter 7: Case Study (11 pages, 63 KBytes)
  • Chapter 8: Conclusion and Summary (3 pages, 13 KBytes)
  • Appendix A (12 pages, 77 KBytes)
  • Appendix B (3 pages, 33 KBytes)
  • Bibliography (9 pages, 118 KBytes)

Abstract

We have developed an agent-based methodology that supports the design of real-time decision-making software in embedded systems. In particular, this thesis addresses the issues of time-bounded decision-making and ease of development and maintenance.

The component-based design involves distributing decision-making among several software agents. Each agent can be viewed as a software module that reacts to changes in its environment. This reactive design leads to deterministic and bounded execution times, which are highly desirable in real-time applications.

Our framework reduces development costs and enhances maintainability by enforcing component-based design. The modular architecture, that encompasses the components, supports partial development and incremental enhancement of functions. We have also devised guidelines for decomposition of a rule base into several decision-making agents. In addition, to integrate and manage components, we have designed a state-based communication mechanism and a control mechanism.

A decision-making solution intended for embedded systems is constrained by resource limitations, including physical size, cost, power consumption, memory, and processing power. The methodology presented in this thesis accounts for these constraints by reducing memory and processing power requirements.

The key contributions of this work include: a software framework that supports distributed problem solving in decision-making systems, a software model for decision-making agents, and a state-based blackboard mechanism for time-bounded communication. The framework includes a control mechanism for managing agents, and a time-bounded mechanism for state-based communication in embedded systems that exhibits low overhead, predictability, dynamic binding, and high reliability.