A Petri net-occam based methodology for the development of dependable distributed control software.

Analysis of flexible manufacturing cells (FMCs) shows their requirement for flexible, correct, reliable, safe and distributed control. A comparison of the state of the art in software engineering for parallel systems, and an examination of safety related systems, reveal a need for formal and rigorous techniques at all stages in the software life cycle. However, parallel software, safety related software and formal techniques are complex. It is better to avoid faults rather than eliminate or tolerate them, and although less flexible, avoidance is often simpler to implement. There is a need for a tool which overcomes many of these complexities, and this thesis discusses and defines such a tool in the form of a methodology. The novelty of the work is in the combination of the core goals to manage these issues, and how the strategies guide the user to a solution which will not deadlock and which is comprehensible. Place-transition Petri nets are an ideal representation for designing and modelling the interaction of concurrent (and distributed) processes. Occam is a high level real time parallel language designed to execute on one or a network of transputers. Transputers are processing, memory and communication building blocks, and, together with occam, are shown to be suitable for controlling and communicating the control as the DCS in FMCs. The methodology developed in this thesis adopts the mathematically based tools of Petri nets, occam and transputers, and, by exploiting their structural similarities, incorporates them in a steps and tasks to improve the development of correct, reliable and hence safe occam code. The four steps: identify concurrent and sequential operations, produce Petri net graphs for all controllers, combine controller Petri net graphs and translate Petri net graphs into occam; are structured around three core goals: Petri net/occam equivalence, comprehensibility and pro-activity; which are manifest in four strategies: output-work-backwards, concurrent and sequential actions, structuralise and modularise, and deadlock avoidance. The methodology assists in all stages of the software development life cycle, and is applicable to small DCSs such as an FMC. The methodology begins by assisting in the creation of DCS requirements from the manufacturing requirements of the FMC, and guides the user to the production of dependable occam code. Petri nets allow the requirements to be specified as they are created, and the methodology's imposed restrictions enable the final Petri net design to be translated directly into occam. Thus the mathematics behind the formal tools is hidden from the user, which should be attractive to industry.The methodology is successfully applied to the example FMC, and occam code to simulate the FMC is produced. Due to the novelty of the research, many suggestions for further work are given