Design-Space Reduction for Architectural Optimization of Automotive Embedded Systems

A key decision for the synthesis of automotive embedded systems is the allocation of application software components to ECUs. Design Space Exploration (DSE) supports the decision by automatically characterizing and evaluating a large number of possible design alternatives, and thereby suggesting the optimal ones. A primary challenge for applying DSE methods to support this decision is to reduce the computation time of the DSE process while maintaining the generality and optimality. This paper exploits legacy system architectures and the AUTOSAR standard to preemptively reduce the design space, because both artifacts limit the flexibility of certain design variables. A new DES formulation incorporating the constraints of the legacy system architectures and the AUTOSAR standard is proposed in this paper. Computation result shows a large reduction of the computation time comparing to traditional modeling and formulations. The scalability of our method is also analyzed by testing it on a set of random problem instances.