This paper describes the foundations for a methodology that may be used in developing products and processes of dynamically evolving, large-scale complex systems (DELCS). Having identified incompleteness in Traditional Systems Engineering's (TSE) historically reductionist, machine-model approach, this methodology represents an alternative form of systems engineering that uses a technique of complexification in ameliorating many problematic effects in the design of complex systems.

Paradoxically, we specifically utilize these complexity characteristics as enabling agents in providing more robust solutions to complex problems in large-scale systems - methods that have historically been considered antithetical to the practice of systems engineering and avoided. We term this new methodology Complex Systems Engineering (CSE).

The guiding principles for the foundations of CSE are:

1. To model the effects of energy on system dynamics, information, entropy and time.

2. To consider both temporal dynamics and state models through dynamical architecture.

3. To facilitate better congruence with scientific foundations.

4. To facilitate reduction in risk from scientific uncertainty.

5. To provide better navigation and visualization.

The method's motivations originated in the domains of systems and software engineering, but are realized through concepts in complexity theory and complexity science as extended from quantum physics, thermodynamics and statistical mechanics, graph theory and inverse theory. These are implemented as heterotic network models embedded in n-dimensional context manifolds extended in temporal dimensions. The network graphs serve as knowledge models that are used to encode, describe and report information for analysis, to simulate behavior, and to provide insight into alternative patterns. Specifically, the methodology searches the large-scale networks for small-world properties, using multiple dimensions of self-similarity in discovering navigational paths and distances. It 'simplifies' complexity, not by reduction but through resolution by adding these discoveries as small functional parameters into the network structure genotype. Therefore it may be described a complex meta-system that replicates complex system models using hyper-dimensional NeuroEvolution of Augmenting Topologies (HyperNEAT).

It is proposed that these techniques may be utilized for engineering such diverse complex dynamical systems as large-scale software systems, collaboration networks, internet fact webs, commercial enterprises, the national defense, and even ad hoc teams within these organizations. It is conjectured better results will be seen compared with using either TSE, game theory, forms of regulated self-organization, highly optimized tolerances, or negotiated group consensus, individually.

This paper is available upon e-mail application to Foundations of CSE