Lund Institute of Technology

Modern machinery, such as automobiles, trains and aircraft, are equipped with embedded distributed computer control systems where the software implemented functionality is steadily increasing. The development of such systems, however, is still a relatively new and young discipline. There is consequently a lack of tools, methods and models to support, in particular, the early architectural design stages. Closely associated with the design of a distributed control system are design decisions on the overall system structure, function triggering and synchronization, policies for scheduling, communication and error handling. These parameters to a large extent determine the resulting system timing and dependability, and hence, impact the control application performance and robustness. The primary target for the research described here is the development of executable models to support architectural design. The report describes earlier efforts in this direction including case studies in the DICOSMOS and AIDA projects. Experiences from the studies are discussed including aspects on modelling to support interdisciplinary design, modeling abstraction and accuracy, and tool implementation. A state of the art survey of related modelling efforts reveals that very little appears to have been done in terms of modelling and simulation that involves simulation of the environment with the computer control system. In addition, the treatment of distributed computer systems, redundant systems, and error scenarios in this context appear to be scarce. The possibilities for extending the simulation models to arrive at a ready to use library to support the design of distributed control systems are discussed. This work is already under way. Remaining work includes to verify the developed simulation models, and to evaluate the models and their usage in case studies. One related topic for further work includes integrating the simulation models with the AIDA modeling framework. A longer term aim is that the models and simulation features should form part of a larger toolset supporting also earlier and later development stages, thus providing more comprehensive analysis and synthesis support.