Abstract :
[en] Cyber-Physical System (CPS) are systems piloting physical processes which have become an integral part of our daily life. We use them for many purposes: transportation (cars, planes, trains), space (satellite, spacecrafts), medical application, robotics, energy management, home appliance, manufacturing, and so many other applications. Model-Driven Engineering (MDE) is widely applied in the industry to develop new software functions and integrate them into the existing run-time environment of a Cyber-Physical System (CPS), for instance, the control software for automotive engines, which are deployed on modern multi-core hardware architectures. Such an engine control system consists of different sub-systems, ranging from an air system to the exhaust system. Each of these sub-systems, again, consists of software functions which are necessary to read from the sensors and write to the actuators. In this setting, MBD provides indispensable means to model and implement the desired functionality, and to validate the functional, the non-functional, and in particular the real-time behavior against the requirements. Current industrial practice in model-based development completely relies on generative MBD, i.e., on code generation to bridge the gap between model and implementation. An alternative approach, although not yet used in the automotive domain is model interpretation. In this thesis, in the place of code generation, we investigate the applicability of model interpretation to automotive software development with a help of a control function design. We present the benefits compared to the existing code-generation practice. The control laws of these software functions typically assume deterministic sampling rates and constant delays from input to output. However, on the target processors, the execution times of the software will depend on many factors such as the amount of interferences from other tasks, resulting in varying delays from sensing to actuating. The literature approaches support the simulation of control algorithms, but not their actual implementation. Further in the thesis, we present the CPAL model interpretation engine running in a co-simulation environment to study control performances while taking the run-time delays into account. The main advantage is that the model developed for simulation can be re-used on the target processors. Additionally, the simulations performed at design phase can be made realistic in the timing dimension through the use of timing annotations inserted in the models to capture the delays on the actual hardware. Introspection features natively available facilitate the implementation of self-adaptive and fault-tolerance strategies to mitigate and compensate the run-time latencies. Experiments on controller tasks with injected delays show that our approach is on-par with the existing techniques with respect to simulation. We then discuss the main benefits of our development approach which are the support for rapid-prototyping and the re-use of the simulation model at run-time, resulting in productivity and quality gains. As the processing power is increasingly available with today's hardware, other concerns than execution performance such as simplicity and predictability become important factors towards functional safety objective. The motivation towards predictable execution behavior, we revisited FIFO scheduling with o set and strictly periodic task activations. The execution order in this case is uniquely and statically determined. This means that whatever the execution platform and the task execution times, be it in simulation mode in a design environment or at run-time on the actual target, the task execution order will remain identical. Beyond the task execution order, the reading and writing events that can be observed outside the tasks occur in the same order. This property, leveraged by our MBD environment CPAL design flow provides a form of timing equivalent behavior between development phase and run-time phase which eases the implementation of the application and the verification of its timing correctness. Thus, the proposed development environment facilitates where also the non-experts are able to quickly model and deploy complex embedded systems without having to master real-time scheduling and resource-sharing protocols. In practice, the design of a software component involves designers from various viewpoints such as control theory, software engineering, safety, etc. In practice, while a designer from one discipline focuses on the core aspects of the field, he / she neglects or considers less importantly the other engineering aspects (for instance, real-time software engineering or energy efficiency). This may cause some of the functional and non-functional requirements not to be met satisfactorily. In the thesis, we present a model-driven co-design framework based on the timing tolerance contract to address such design gaps between control and real-time software engineering. The framework consists of three steps: controller design, verified by jitter margin analysis along with co-simulation, software design veri fied by a novel schedulability analysis, and the run-time verification by monitoring the execution of the models on target. This framework builds on earlier mentioned CPAL design environment, which enforces a timing-realistic behavior in simulation through timing and scheduling annotations. Through various case studies, we show that our tool enables not only to automate the analysis process at design time but also to enhance the design process by systematically combining models and analyses.
