Clonal selection algorithm for partitioning and scheduling of codesign systems

In system-level design, applications are presented as task graphs where tasks; i.e. nodes of the graph, have several implementation options differing in some criteria such as time, area and power. Systems designed with this approach are those that are application specific and for performance reasons are implemented in a hardware/software codesign manner. In this paper the most important design issues in these systems i.e. partitioning and scheduling are investigated. Our approach, namely CSPA, is a heuristic algorithm inspired by the biological immune system and attempts to obtain an optimal design for a given system composed of several hardware and software components. We use a graph representation of the system where nodes are operational components and edges are communication links between them. We propose an immune-based approach which based on artificial immune system and apply the clonal selection algorithm as one of the different types of algorithms inspired by biological systems. To date there is no work in this field that uses the clonal selection algorithm for optimization of partitioning. Empirical results show a suitable improvement by using this approach in comparison with traditional evolutionary algorithms and also traditional immune-based approach.