A multi-layer detailed routing approach based on evolutionary algorithms

Abstract

We present an evolutionary algorithm (EA) for detailed routing problems (DRPs), like the channel routing problem and the switchbox routing problem. We combine EAs with domain specific knowledge, i.e. the genetic operators make use of the rip-up and reroute technique. The algorithm can work with two layer and multilayer problem instances. The efficiency of our algorithm is demonstrated by application to multilayer channel routing benchmarks. Instances with up to five layers, 130 columns, and more than 60 nets are considered.