1. Introduction

The use of models has found widespread application in systems engineering as a (semi-)formal method to manage the complexity and heterogeneity of large scale systems and their design teams and tools because they are amenable to analysis and synthesis tasks. For example, models are used for knowledge representation, requirements engineering [48,63], structured analysis [24,61], to manage complexity and achieve high quality of engineered systems [62], to handle the heterogeneous nature of embedded systems [21], as a high level programming method [23,22,55], and to bridge conceptual differences between domains [54]. With the advent of ubiquitous computing, model-based applications, e.g., in control, diagnosis, and maintenance, will become pervasive and ultimately become as proliferated as embedded computing power [35,12]. To avoid overspecification and attain optimal performance, new design paradigms based on holistic views (e.g., mechatronics [57,56]) are a necessity to analyze subtle interaction between information processing components and the physical environment as well as between the different design tasks. This requires tight integration of the separate individual design activities. However, each of the engineering disciplines involved in system design and operation have developed domain and problem specific (often proprietary) formalisms that match their needs optimally but complicate the integration process. The goal of this research is to develop and prototype a core of next generation multi-paradigm modeling methods and technologies that address this incompatibility and enable the development of novel applications. This is a powerful approach that allows the generation (instantiation) of domain and problem specific methods, formalisms, and tools and because of a common meta language, these different instances can be integrated by combination, layering, heterogeneous refinement, and multiple views [18,60]. At present this is still very much a topic of on-going research that breaks down into two types of activities: (i) heterogeneous modeling [14,28,31,47] and formalism [6] and tool coupling [15,16], and (ii) behavior generation [7,29,36,46,51,58]. The first is mainly concerned with the symbiotics (symbols, syntax, and static semantics) of modeling formalisms, whereas the second addresses analysis and behavior generation using the dynamic semantics of such heterogeneous models. Important issues include but are not limited to the design of system engineering ontologies [8], integrated development environment design, heterogeneous execution models [34], code synthesis (software and hardware description language) [49,25,32], and formal methods [11]. Three orthogonal dimensions of multi-paradigm modeling are multi-abstraction modeling, concerned with the relationship between models at different levels of abstraction, possibly described in different formalisms, multi-formalism modeling, concerned with the coupling of and transformation between models described in different formalisms, and meta-modeling, concerned with the description of model representations and instantiation of domain specific formalisms [45]. When extended with sophisticated model transformation facilities, the multi-paradigm modeling notions can be exploited to facilitate a suite of technologies and applications that manipulate a model into a different representation, possibly changing the abstraction, partitioning, and hierarchical structure to render it suitable for particular tasks, i.e., it is operated on the model rather than its generated information. Though some model transformation schemes exist within [13,27,33,59] and between formalisms [4,5,53,58], a hiatus in this multi-paradigm modeling effort is the prevalent need to manually design models in different representations for analyses, consistency checks, and execution. The model transformations that are available and current development efforts tend to focus on the goal of system realization from design (e.g., automatic code synthesis) while models embody knowledge, and as such they also form the core of intelligent applications (e.g., model-predictive control [1], model-based diagnosis [10,26,37,39,40,43], and self maintenance). When extended with sophisticated model transformation facilities, the multi-paradigm modeling notions can thus be exploited further to facilitate a suite of technologies and applications that implement a form of higher intelligence: Where present intelligent applications utilize a formal representation of some form of a process or system to derive information about its state and predict future behavior, higher intelligence manipulates this model into a different representation, possibly changing the abstraction, partitioning, and hierarchical structure to render it suitable for required tasks, i.e., it operates on the model rather than its generated information.