Research Of Model Transformation Based On Real-Time Semantic Model And Semantic Consistency

Such typical complex industrial control systems as industrial robots and CNC machines are expected more real-time constraints, higher concurrency and multi-point collaboration, so the design and implement of accurate and efficient real-time control software is very important. In the process of real-time control software design and development, it will be extremely important to transfer the design model with both control functions and non-functional conventions into an executable implementation model for specific target platform truthfully and completely. Design model and implementation model are different in forms of expression. Except carrying out the control algorithms and control laws correctly, the implementation model must accurately identify the Non-functional semantic constraints of the design model and maintains them, such as timing semantics and interacting parallel behavior semantics. Therefore, how to keep the functional integrity and ensure the semantic consistency in model transformation is particularly crucial.The industrial real-time control domain has some Special performances when using model-driven development method. Generally real-time control software is executed with concurrent processing multitask. Concrete modeling for a large number of communications and collaborations is the main job in the process of design and development, so there are strictly requirements for behavior timing and as a result the semantic architecture is complex. In order to get more effectively source model, it is necessary to increase the independence and the degree of abstraction of the source model. In order to get more practical target model, it is necessary to increase the dependence and the specific degree of the target model. As a result this will increase the gap both in form representation and in semantic. To satisfy the accuracy, integrity and semantic consistency in the model transformation we need the help of a good model transformation methods and techniques.Current researches focus on concrete model transition technology between different abstract layers, and they pay more attention to the model framework mapping rather than specific actions of mapping. They can not support the real-time control software design expectance and semantic consistency maintenance requirement for behavior modeling and conversion. The existing model transformation methods have the following problems:1) they transfer the source model directly to the target model so there will exist wider semantic gap and will introduce larger semantic divergence; 2)they do not support the model construction and transformation of concrete behavior, so they lack of precise semantic definition; 3) they do not support the synchronous construction and transformation of timing and function, so they can not ensure the integrity of semantic mapping; 4) they can not generate the target model which are demanded by real-time control system platform, so they can not guarantee the semantic consistency in the process of model transformation of real-time control system. Therefore, it is difficult for them to fulfill the model transformation from the platform independent model to the platform specific model in real-time control software MDD development.This paper attempts to research the model transformation problem in the domain of real-time control system development. The target of this research is to propose a kind of model transformation method especial for real-time system which can solve the semantic consistency maintenance problem in the conversion process. The main idea is to construct real-time semantic model based on real-time ontology semantic knowledge base, which act as intermediate between the source model and the target model. This research provides an effective support scheme for the semantic keeping in the model transformation.The main contents and contributions of this thesis are as fellows.Firstly, real-time semantic model is defined and a model transformation method based on real-time semantic model is supplied. Real-time semantic model is defined as the middle layer between the source model and the target model. Differenting from source model in graph representation. Real-time semantic model is defined with formal meta-language, so it is a platform independent model. At the meantime real-time semantic model is the abstraction of target model whose semantic edifice is described with ontology language. The definition, rule, timing and resource in the semantic system are extracted from target platform directly. The model transformation includes two steps:source model-semantic model transition and semantic model-target model transition. The former completes the definition mapping, rule mapping and timing mapping; the latter fulfils the type mapping, structure mapping and resource semantic mapping. By this means, the semantic integrity and semantic consistency are guaranteed.Secondly, an ontology framework for multi-task interactive based on real-time semantic model is built. The ontology knowledge system of real-time systems is established by means of identification and interpretation for the ontology concept, and the multi-task interactive ontology application framework based on real-time semantic model is presented. The conception of object, behavior and programmed algorithm are described and the meta-language definitions of which are supplied respectively. Especially, the decomposition and refinement of behavioral model are studied as emphasis and an "object-related" behavior model classification method is proposed. The timing semantic is expressed by prefix notation and sequential operator.Finally, the model mapping based on real-time semantic model and semantic consistency holding are realized. The mapping principle from source model to semantic model is generalized. The static semantic consistency holding method based on ontology knowledge base is realized. The timing semantic in the behavior mapping is keeping by event structure. The mapping principle from semantic model to target model is generalized. Type mapping, structure mapping and resource semantic holding are realized and finally a model mapping scheme from semantic model to iRMX target model is given.