A parallel implementation of evolutionary divide and conquer for the TSP

Evolutionary divide and conquer (EDAC) applied to the geometric travelling salesman problem uses a genetic algorithm to explore the space of problem subdivisions. The underlying techniques for subdivision and patching are based on the cellular dissection algorithms of Richard Karp (1977), but the addition of new repair heuristics as well as the genetic algorithm, appears to have succeeded in lifting the quality of solution way above Karp's algorithms, whilst maintaining almost linear scaling of the algorithm with increased problem size. In this paper we outline a parallel implementation of EDAC and present some preliminary results of this work.