The Uncertainty Research In Requirements Engineering

It is the most difficult to determine the aim of the software system in software development Traditional requirement analysis depends on developers' understanding of users and their own context knowledge to obtain a certain and invariable requirements specification. However, because users' requirements are usually dependent on environment and uncertain, certain specification can not reflect users' real requirements and their dynamic though it benefits succeeding design. So requirements engineering is presented. People hope to solve all kinds of uncertain problems by systematic and engineering methods, hi fact, requirements engineering is the combination of formal methods and informal methods as well as the bridge of users and software engineers.It is necessary and significant to introduce the methods of flexibility logic and artificial intelligence into the research on requirements engineering. The aim is to induce the model and the norm form of the problem by engineering idea and formal method according to the reality condition and concrete demands and to make them include various uncertainty when users present informal requirements, as is advanced task and difficulty in the present research on software engineering.hi fact, uncertainty problems in requirements engineering are about translating users' requirements into goals of the system. Uncertainty mainly includes imprecision and inconsistency, the latter of which can be divided into static inconsistency and dynamic inconsistency. Traditional fuzzy logic and default logic can not dynamically and completely describe uncertain requirements of users. Universal logics is full azimuth, flexible and opening logics. It supplies better formal basis for both uncertainty requirements. This thesis is based on universal logics. It researches on uncertainty in requirements engineering from generalized fuzzy logic operators and variable default theory, establishes a flexible reasoning mechanism and discusses default rules and domain mapping around the generation and maintenance of requirements specification.The main contributions and creative contents are as follows:1. This thesis deals with imprecise requirements based on universal logics, establishes flexible frame based on extensive pertinence to describe relations among imprecise requirements, which means describing AND, OR, IMPLICATION and EQUIVALENCE by continuous families of operators. The revision to requirements specification is different from traditional belief revision in at least two aspects. One is that inconsistent requirements can cooperate, so cooperation under different conditions can be realized by continuous families of averaging operators, the other is that there are differences among requirements, namely, main body of requirements can be regarded as knowledge (fact) and the complement as belief, setting threshold to realize dynamic translation between them according to experience.2. Present change-driven requirements principle. We think that inconsistent requirements are inevitable, which can only be reasoned under inconsistent and evolutiveenvironment. This thesis applies default theory which is adaptive to change to the domain of requirements engineering, introduces partial requirements specification (requirements model), which helps to avoid trivial solution and obtain locally effective result whose correctness can not be absolutely guaranteed. It is similar to our method that people deal with incomplete information world by commonsense.3. Since the greatest difficulty of this formal method is to make imprecise requirements hypothesis space more reasonable, this thesis presents a method of determining background knowledge by mapping knowledge, which can not only supply acceptable formal specification that accords with commonsense, but also implement requirements reuse among domains.4. The acquirement of incomplete knowledge is key to adapt variable incomplete information environment. This thesis discoveries default rules via generalized similarity relation, which...