Research On Requirement Traceability Of The Safety-critical Systems

Embedded system is widely used in aviation,aerospace,transportation,and other important fields which usually has a high demand of reliability and security.Once these systems become disabled which will lead to significant damage to property,environmental damage and casualties and other major consequences.Software engineering for safety-critical systems is particularly difficult.One of the difficulty is addressing any legal and regulatory requirements.For the safety and reliability of safety-critical system,lack of requirement errors and faults plays a decive role during the development process.Thus,building the traceability among the artifacts of safety-critical system is a very important procedure,especially the relationship between requirement and design model.Requirement is now mostly described by natural language,or methods like RUCM which is semi-formal method to describe system requirements.Meanwhile,formal or semi-formal methods are usually used to design architecture models,which leads to the difficulty to ensure the consistency between requirement and architecture model of the safety-critical system.According to the statistic of software projects: more than 50% of the system error is caused by error requirements or lack of requirements,and more than 80% of the cost spent on tracking these requirement errors.These errors can be discovered and fixed earlier by improving requirements specification and design.Thus,one can see that reduction of the errors in requirement and building traceability between requirement and models are important for developing a reliable software.In the thesis,Complex embedded real-time system architecture modeling language AADL(Architecture and Analysis the Design Language)used as architecture design language,can effectively describe the runtime system architecture model.AADL owns sub language annex such as error model annex that makes AADL model the system better.AADL requirement specification gives out a assurance framework called ALISA(Architecture-Led Incremental System Assurance)which is a method used to check whether design of the model can meet the corresponding requirement.At present,there are still many problems in using ALISA to build the traceability betwee natural language requirement and AADL model.ALISA provides a formal requirement specification but there is still a large gap between the ALISA requirement specification and natural language requirement.So natural language requirement cannot be transformed to ALISA requirement specification automatically.ALISA verification framework focuses on a particular property of a component of a particular AADL model.Due to AADL model owning several abstract levels,several components and many properties.As a result,checking the AADL model manually may need much work.Another problem is that ALISA only considers that checking the AADL core language instead of taking AADL error model annex into consideration.To solve the problems above,this thesis proposes a method which utilizes the ALISA framework to achieve checking the consistency between static semantic of AADL model and the corresponding natural language requirement.The main research content is as follows:(1)For the ambiguity and uncertainty problem of natural language,a domain-specific restricted natural language requirement template is designed to describe the requirement of the safety-critical system.Then the requirement template can be transferred to ALISA formal requirement specification by the way of model transformation.(2)For the problem that AADL multilayer design model is hard to be verified manually,according to the ALISA requirement specification and AADL model,using ATL(ATLAS Transformation Language)technology to define transformation rules via the meta models.Then ALISA verification plan can be automatically generated which is used to check key values or important attributes in AADL model.(3)For the problem that AADL error model annex cannot be described by ALISA requirement specification,this thesis extend the requirement specification to describe the error states,and connect it with the elements of ALISA framework.At last,on the basis of the error model and requirement specification,generate verification plans automatically via the ATL techology.According to the method proposed above,this thesis designs and develops the tools and conduct an analysis on the thermostat system of an isolette(infant incubator).On the basis of the definition of consistency,engineers can judge if the system design meets the requirement to ensure requirement earlier during the development process and reduce the cost.Thus,the effectiveness and feasibility of the method in this thesis can be revealed.