An O(N log N) expected time merge heuristic for the planar ETSP

The Euclidean Traveling Salesman Problem (ETSP) may be stated as follows: if we have a set of N cities and a salesman has to visit each of the N cities exactly once and then return to his starting point, what is the order in which he should visit the cities (called the tour) in order to minimize the total distance which he has to travel? In this problem, the cities are given as a set of points in the Euclidean plane, i.e, they are given in cartesian coordinates, and the distances between the cities are measured using the Euclidean measure of distance in the plane. Since this problem is NP-complete, much attention has been focused on heuristics which run in polynomial time and generate tours which are close to optimal. In general, there is a tradeoff between speed and quality; i.e, faster algorithms do not approximate the optimal tour as well as some costlier algorithms. Thus it is necessary to achieve a suitable balance between cost and quality.