No description available.

No description available.

This thesis extends the body of research into the application of Genetic Algorithms to the Project Scheduling Problem (PSP). A thorough literature review is conducted in this area as well as in the application of other similar meta-heuristics. The review extends previous similar reviews to include PSP utilizing the Design Structure Matrix (DSM), as well as incorporating recent developments. There is a need within industry for optimisation algorithms that can assist in the identification of optimal schedules when presented with a network that can present a number of possible alternatives. The optimisation requirement may be subtle only performing slight resource levelling or more profound by selecting an optimal mode of execution for a number of activities or evaluating a number of alternative strategies. This research proposes a unique, efficient algorithm using adaptation based on the fitness improvement over successive generations. The algorithm is tested initially using a MATLAB based implementation to solve instances of the travelling salesman problem (TSP). The algorithm is then further developed both within MATLAB and Microsoft Project Visual Basic to optimise both known versions of the Resource Constrained Project Scheduling Problems as well as investigating newly defined variants of the problem class.

No description available.

No description available.

In recent years, simulation modelling of software processes have has promoted as a tool to understand, study, control, and manage software development processes. Claims have been made that simulation models are useful and effective at gaining insight into software development processes. However, little has been said about the process of developing simulation models for software engineering problems. Simulation modelling is a young discipline in software engineering. Consequently, many number software process simulation modellers are thought to be novices. The simulation modelling process is believed to have had an effect on the quality of a simulation study. Although there is a body of knowledge available in the general simulation literature to guide and educate novices, the software process simulation modelling literature lacks information for novice software process simulation modellers to understand and adopt a simulation modelling process. This thesis aims to develop a simulation modelling process for novice software process simulation modellers. This thesis reports how the development and evaluation of a simulation modelling process for novice software process simulation modellers. The rapid simulation modelling process (RSMP) is based on an empirical study of the contexts and practices of expert simulation modellers in SPSM and Operational Research (OR). The RSMP is intended to be independent of a particular simulation technique (i. e. system dynamics or discrete event simulation) and guides novice software process simulation modellers through a set of steps that should be undertaken during a simulation study; the RSMP emphasises heavy client contact and provides guidelines for model documentation. The RSMP has been evaluated through controlled experiments with novice software process simulation modellers using system dynamics (SD) modelling. In the future, it will be further evaluated with software process simulation modellers using discrete event simulation. The RSMP has also been evaluated with a panel of expert software process simulation modellers. The main contribution of this study lies in providing novice software process simulation modellers with a simulation modelling process, which embodies real world simulation practice and is intended to be independent of a particular simulation technique.

No description available.

No description available.

In software or hardware development, we take an abstract view of a process or system - i.e. a specification - and proceed to render it in a more implement able form. The relationship between an implementation and its specification is characterised in the context of formal verification using a notion called refinement: this notion provides a correctness condition which must be met before we can say that a particular implementation is correct with respect to a particular specification. For a notion of refinement to be useful, it should reflect the ways in which we might want to make concrete our abstract specification. In process algebras, such as those used in [28,50,63]' the notion that a process Q implements or refines a process P is based on the idea that Q is more deterministic than P: this means that every behaviour of the implementation must be possible for the specification. Consider the case that we build a (specification) network from a set of (specification) component processes, where communications or interactions between these processes are hidden. The abstract behaviour which con- stitutes these communications or interactions may be implemented using a particular protocol, replication of communication channels to mask possible faults or perhaps even parallel access to data structures to increase perfor- mance. These concrete behaviours will be hidden in the construction of the final implementation network and so the correctness of the final network may be considered using standard notions of refinement. However, we can- not directly verify the correctness of component processes in the general case, precisely because we may have done more than simply increase determinism in the move from specification to implementation component. Standard (pro- cess algebraic) refinement does not, therefore, fully reflect the ways in which we may wish to move from the abstract to the concrete at the level of such components. This has implications both in terms of the state explosion prob- lem and also in terms of verifying in isolation the correctness of a component which may be used in a number of different contexts. We therefore introduce a more powerful notion of refinement, which we shall call refinement-after-hiding: this gives us the power to approach ver- ification compositionally even though the behaviours of an implementation component may not be contained in those of the corresponding specification, provided that the (parts of the) behaviours which are different will be hidden in the construction of the final network. We explore both the theory and practice of this new notion and also present a means for its automatic verifi- cation. Finally, we use the notion of refinement-after-hiding, along with the means of verification, to verify the correctness of an important algorithm for asynchronous communication. The nature of the verification and the results achieved are completely new and quite significant.

For an engineer to be able to modify successfully a complex computer-based system, he will need to understand the system's functionality. Traceability can help the engineer to gain that understanding, but several surveys have observed that traceability information is poorly recorded. This thesis argues, based on a survey of nine aerospace projects, that one of the main causes of poor recording is that Traceability does not directly benefit the development process. The recording of traceability information is best performed by the engineers directly involved in the development process, yet it is precisely these engineers who seem to obtain no direct benefit in performing this task. This can be summarised as the Traceability Benefit Problem. To overcome this problem the recording of traceability data must provide immediate, tangible benefits to the engineers involved in the current development process. A related problem that occurs in large multi-team projects that follow development processes based on predictive models (such as Waterfall or VModel) is the changing of interface documentation without adequate negotiation (referred to as Throwing the Problem over the Wall). This thesis describes, in detail, how a small automotive sensor project addressed these problems by developing a Requirements Traceability system that enabled the reuse of software and provided a basis for the negotiation of changes with their customer. Analysis of the lessons learnt from the automotive sensor and aerospace projects lead to the definition of the Traceable Development Contract. The contribution of this thesis is the description and discussion of the Traceable Development Contract, a method of coordinating the interaction of related development teams in development process that is based on a predictive development model. The Traceable Development Contract is proposed as a means of controlling the upstream team bias with respect to the imposition of changes, by employing traceability to provide a basis for the negotiation of change. By VI employing traceability in this way, it becomes beneficial to the development engineers and therefore overcomes the Traceability Benefit Problem. Finally, the thesis considers how the Traceable Development Contract traceability information can be exploited further to provide solution maturity and design metrics.