Directed Convergence Heuristic: A fast & novel approach to Steiner Tree Construction

One of the fundamental problems encountered during the VLSI design flow is to find minimum length nets that connect specific nodes on the chip. The challenge lies in finding an efficient solution to the Steiner tree problem in graphs (SPG) that not only maximizes the routing efficiency but also lends itself well for fast implementation. In this paper, we propose a new and innovative approach for solving the Steiner tree problem, called the "directed convergence heuristic (DCH)". In essence, the DCH based algorithm places entities called pawns on nodes that need to be connected. These pawns move towards each other in a directed fashion and while doing so, leave trails for constructing a Steiner tree between the nodes. When all the pawns converge at a node, the trails merge to create a Steiner tree. Experimental results on benchmark Steiner trees show that DCH is robust and converges faster while yielding competitive near optimal solutions. The algorithm is amenable for implementation on parallel computing architectures