Abstract:The Traveling Salesman Problem (TSP) in the two-dimensional Euclidean plane is among the oldest and most famous NP-hard optimization problems. In breakthrough works, Arora [J. ACM 1998] and Mitchell [SICOMP 1999] gave the first polynomial time approximation schemes. The running time of their approximation schemes was improved by Rao and Smith [STOC 1998] to $(1/\varepsilon)^{O(1/\varepsilon)} n \log n$. Bartal and Gottlieb [FOCS 2013] gave an approximation scheme of running time $2^{(1/\varepsilon)^{O(1)}} n$, which is optimal in $n$. Recently, Kisfaludi-Bak, Nederlof, and Węgrzycki [FOCS 2021] gave a $2^{O(1/\varepsilon)} n \log n$ time approximation scheme, achieving the optimal running time in $\varepsilon$ under the Gap-ETH conjecture. In our work, we give a $2^{O(1/\varepsilon)} n$ time approximation scheme, achieving the optimal running time both in $n$ and in $\varepsilon$ under the Gap-ETH conjecture.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to provide an approximate algorithm scheme that is optimal in both time and precision for the Traveling Salesman Problem (TSP) in the two - dimensional Euclidean plane. Specifically, the authors aim to design an approximate algorithm with a running time of \(2^{O(1/\varepsilon)}n\) to achieve the optimal dependence on both \(n\) and \(\varepsilon\) simultaneously, and it is optimal under the Gap - ETH conjecture.
### Problem Background
The Traveling Salesman Problem (TSP) is a classic NP - hard optimization problem, whose goal is to find the shortest traversal path among a given set of \(n\) points. For the TSP in the two - dimensional Euclidean plane, Arora [J. ACM 1998] and Mitchell [SICOMP 1999] respectively proposed the first polynomial - time approximate algorithms. Subsequently, Rao and Smith [STOC 1998] improved its running time to \((1/\varepsilon)^{O(1/\varepsilon)}n\log n\). Bartal and Gottlieb [FOCS 2013] proposed an approximate algorithm with a running time of \(2^{(1/\varepsilon)^{O(1)}}n\), which is optimal in terms of \(n\), but not as good as the result of Rao and Smith in terms of \(\varepsilon\). Recently, Kisfaludi - Bak, Nederlof and Węgrzycki [FOCS 2021] proposed an approximate algorithm with a running time of \(2^{O(1/\varepsilon)}n\log n\), which is optimal under the Gap - ETH conjecture in terms of \(\varepsilon\), but still has a factor of \(\log n\) in terms of \(n\).
### Main Contributions of the Paper
The authors of this paper propose a new approximate algorithm, whose running time is \(2^{O(1/\varepsilon)}n\), and it is optimal under the Gap - ETH conjecture in terms of both \(n\) and \(\varepsilon\). Specifically, the authors solve the problem of handling single - crossing boundaries. By introducing the concepts of new port sets and \(r\)-basic tours, the algorithm can achieve a high - precision approximate solution while maintaining a linear - time complexity.
### Main Technical Contributions
1. **New Definition of Port Sets**: In order to handle the case of single - crossing boundaries, the authors introduce new port sets, which include the port sets of Bartal and Gottlieb [BG13] and possibly an additional point, that is, the intersection of the 2 - approximate solution \(T\) and the boundary.
2. **\(r\)-basic Tours**: The authors define a new type of \(r\)-basic tours, ensuring that the tour path passes through only pre - defined ports on the boundary of each cell, and each port is visited at most twice.
3. **Structural Theorem**: The authors prove a structural theorem, indicating that there exists an approximately optimal \(r\)-basic tour path.
4. **Dynamic Programming Algorithm**: Based on the above - mentioned structural theorem, the authors use the standard dynamic programming method to calculate the optimal \(r\)-basic tour path, thereby achieving an approximate algorithm with a linear - time complexity.
Through these innovations, the authors successfully reduce the time complexity of the approximate algorithm for TSP in the two - dimensional Euclidean plane to \(2^{O(1/\varepsilon)}n\) and achieve optimality under the Gap - ETH conjecture.