Abstract

Wire routing in a VLSI chip often requires minimization of wire-length as well as the number of intersections among multiple nets. Such an optimization problem is computationally hard for which no efficient algorithm or good heuristic is known to exist. Additionally, in a biobjective setting, the major challenge to solve a problem is to obtain representative diverse solutions across the (near-) Pareto-front. In this work, we consider the problem of constructing spanning trees of two geometric graphs corresponding to two nets, each with multiple terminals, with a goal to minimize the total edge cost and the number of intersections among the edges of the two trees. We first design simple heuristics to obtain the extreme points in the solution space, which however, could not produce diverse solutions. Search algorithms based on evolutionary multiobjective optimization (EMO) are then proposed to obtain diverse solutions in the feasible solution space. Each element of this solution set is a tuple of two spanning trees corresponding to the given geometric graphs. Empirical evidence shows that the proposed evolutionary algorithms cover a larger range and are much superior to the heuristics.

Citations