Clearing the Combinatorial Fog: Tracing the Hidden Paths of TSP Heuristics

Over decades of Traveling Salesperson Problem (TSP) research, powerful heuristics have been developed that efficiently solve many TSP instances. Among them, the local search optimizer LKH and the genetic algorithm EAX stand out as the two complementary state-of-the-art solvers. Yet, the links between instance structures and solver complementarity remain obscure, i.e., it is often unclear how instance structures affect solver performance and behavior.