Mapping functional behavior onto architectural model in a model driven embedded system design

Abstract

The ever-increasing complexity in embedded systems especially in the automobile industry in the recent years has necessitated model-driven engineering. In this paper, we consider the problem of mapping functional behavior onto an architectural model captured in AADL in order to optimize end-to-end delay in executing a distributed function on the specified platform architecture. Our work presupposes that an architectural platform model is fixed due to existing hardware platform that the designers have to work with, whereas a specific functional feature is being designed in software, and implemented on the given platform. We therefore, consider the problem as a behavior modeling followed by mapping of behavioral components including computation, and communication. This mapping requires both spatial mapping as well as temporal mapping. Spatial mapping means binding of computational nodes in the behavioral model to processors/controllers etc., communications to the platform bus; and the temporal binding is done by scheduling the computation on the platform. We explain our method by way of a case study of an adaptive cruise control(ACC) system whose behavior model is represented with a data-flow graph(DFG) captured in LUSTRE. A static schedule is derived from the DFG and then mapped to the platform architecture model by formulating a non-linear optimization problem. The resulting problem being at least NP-hard, we propose use of simulated annealing or other heuristic algorithms to solve the optimization problem. The limitations of our method are discussed as well.