Controllers are in most cases based on periodic sampling and assume a negligible or constant latency between input and output (sampling and actuation). This is something that in many cases can be difficult or costly to achieve. Time-triggered solutions based on static scheduling are one solution, but are sometimes too inflexible or are incompatible with the rest of the system software. In event-based solutions, pre-emption, blocking, execution time variations and non-deterministic kernels generate sampling jitter and latency jitter. The same holds for event-based network protocols. The problem can be approached in different ways. Robust design can be applied to guarantee a certain level of temporal robustness. Techniques can be used to compensate for the timing variations, either passively based on off-line information about the characteristics of the variations, or actively using measurements. The interaction between control and real-time computing becomes extra important in situations where the computing and communication resources are severely limited, e.g., in embedded control system applications, where separation of concerns-based design principles, with strict interfaces between control and computing, may be unfeasible. Instead it is necessary to take both computing and control aspects into account simultaneously. This requires theory and tools that support codesign. From a pure real-time systems approach it is also desirable to provide more flexible ways of allocating computing resources to different applications or tasks. The area of adaptive or feedback-based resource scheduling is one example of this.

Taking implementation issues and limited resources into account in the control design is covered by the terms resource-aware control and implementation-aware control. The development in this area needs to be matched with the similar developments within the real-time field. It is necessary to create models of computation and scheduling, and system software and hardware, which are tailored to the true needs of control applications. This is covered under the terms of control-aware computing and networking.

Model integration and management constitute key challenges in the design of embedded systems; this is also relevant for embedded control systems. Consider for example the design of an embedded automotive ABS braking system. One obvious concern is that of the core motion control functionality, especially the control logic and algorithms and the dynamic behaviour of the system. However, this is only one out of several aspects. Other aspects include safety, security, network communication, mechanical design, IO, power, etc. These aspects and components are in addition typically handled by different specialists, employing different modeling languages and tools, and moreover belonging to different organizational entities. There is therefore a strong industrial need for solutions that support model and tool integration, as well as model management. There are several related research issues including model transformations and methodology. Some confusion is inevitable in this area since it is approached from so many directions (engineering disciplines, information management, tool specific solutions, standardization etc.). We believe that establishing modelling frameworks that characterize the problem and solution space are very important for the progress of the area. An initial such framework has been proposed by the cluster participants (see http://www.md.kth.se/RTC/ARTIST2/publications/CACSD06_Chen.pdf)
A new area is control of computer software systems, e.g., large eCommerce servers. These servers are complex dynamic systems with high levels of uncertainty. The need for control arises at several levels, e.g., admission control, delay control, and utilization control. Several new challenges apply. Since the servers are engineering artefacts, first principles do not apply, at least not on the macroscopic level. Several competing modeling formalisms needs to be combined, e.g., continuous-time flow models, queuing models, and various types of event-based models. System stability has an unclear meaning, and the whole issue of how to write controllable and observable software system is still largely unexplored.

Réalisation Axome - Création de sites Internet