An Improved Integrality Gap for Asymmetric TSP Paths

The Asymmetric Traveling Salesperson Path Problem (ATSPP) is one where, given an asymmetric metric space $(V,d)$ with specified vertices s and t, the goal is to find an s-t path of minimum length that passes through all the vertices in V. This problem is closely related to the Asymmetric TSP (ATSP), which seeks to find a tour (instead of an $s-t$ path) visiting all the nodes: for ATSP, a $\rho$-approximation guarantee implies an $O(\rho)$-approximation for ATSPP. However, no such connection is known for the integrality gaps of the linear programming relaxations for these problems: the current-best approximation algorithm for ATSPP is $O(\log n/\log\log n)$, whereas the best bound on the integrality gap of the natural LP relaxation (the subtour elimination LP) for ATSPP is $O(\log n)$. In this paper, we close this gap, and improve the current best bound on the integrality gap from $O(\log n)$ to $O(\log n/\log\log n)$. The resulting algorithm uses the structure of narrow $s$-$t$ cuts in the LP solution to construct a (random) tree spanning tree that can be cheaply augmented to contain an Eulerian $s$-$t$ walk. We also build on a result of Oveis Gharan and Saberi and show a strong form of Goddyn's conjecture about thin spanning trees implies the integrality gap of the subtour elimination LP relaxation for ATSPP is bounded by a constant. Finally, we give a simpler family of instances showing the integrality gap of this LP is at least 2.