Robust optimization of SoC architectures: A multi-scenario approach

Multimedia and computational intensive applications are more and more pushing towards high-performance, low-power consumption and limited area occupation making the concept of optimality multi-objective. In the context of a design space exploration framework to support the platform-based design approach, we address the problem of robust optimization of a parameterized system-on-chip platform towards a set of objective functions. The concept of robustness deals with the uncertainty of the scenarios for which the system is optimized. The problem of the robust optimization has been treated as a minimization of a geometric mean of the objective functions computed over each uncertainty scenario and it is solved with a multi-criteria decision technique. The resulting robust solution is characterized by a multi-objective optimality over the set of considered scenarios. Experimental results have been gathered from two virtual platforms in a multi-scenario approach.