Software architects are increasingly relying on self-organizing mechanisms to design distributed systems within a dynamic, noisy and unpredictable envi- ronment. At any point in time, no centralized entity has complete knowledge of the state of the environment as a whole and self-organizing mechanisms are mainly naturally-inspired which enables the decentralized control. There are two tracks that drive the research in this thesis: first, as it is the case with any new software engineering paradigm, the successful and widespread deployment of self-organizing systems require notations that explore the use of self-organizing related abstractions and promote the traceability from the design models to code, and engineering methods that provides know-how and guides an engineer during an application design. The second research track is to promote software reuse of self-organizing systems. This thesis’s goals are: to provide a simulation-based engineering method to support the design, development, simulation, validation and refinement of self-organizing multi-agent systems; and to provide a simulation-based architecture. An ar- chitectural design helps on the development of a modular program structure and on the representation of the control relationships between modules and encourages the software engineer to concentrate on architectural design be- fore worrying about optimizations or code. We present: an iterative develop- ment life-cycle based on a customization of the Unified Process, the SSOA – Simulation-based Self-Organizing Architecture –, and the framework that implements the architecture. We have evaluated the design method and ar- chitecture using three different application domains: the automated guided vehicles, the contracts that govern emergent multi-agent systems, and the stem cell behavior computational modeling.