Combinatorial Optimization for Intelligent Control Systems

Combinatorial problems that have to be solved in modern intelligent control systems frequently possess exponential worst-case complexity. This fact precludes the solution of many practical problems with conventional processors. Specialized hardware based implementations are also not viable, in particular because of the inherent heterogeneity of combinatorial problems. Reconfigurable devices offer an alternative solution, which can be customized to the requirements of a specific algorithm and reutilized for other algorithms via a simple reprogramming of their internal structure. In this paper different ways of designing reconfigurable combinatorial accelerators are explored. 1