The major objective of this research is to propose a design architecture for autonomous defense systems for supporting highly intelligent behavior by combining decision, perception, and action components. Systems with such high levels of autonomy are critical for advanced battlefield missions. By integrating a plenty of advanced modeling concepts such as system entity structure, endomorphic modeling, engine-based modeling, and hierarchical encapsulation & abstraction principle, we have proposed four layered design methodology for autonomous defense systems that can support an intelligent behavior under the complicated and unstable warfare. Proposed methodology has been successfully applied to a design of autonomous tank systems capable of supporting the autonomous planning, sensing, control, and diagnosis.