[Paper Review] The guillotine approach for TSP with neighborhoods revisited

Sophie Spirkl|arXiv (Cornell University)|Dec 2, 2013

Advanced Graph Theory Research8 references3 citations

TL;DR

This paper revisits Mitchell's guillotine subdivision approach for the Euclidean TSP with neighborhoods (TSPN), correcting critical gaps in the original PTAS claim for disjoint unit disks and proving a (1+O(ε))-approximation via a refined grid-rounded, (m,M)-guillotine dynamic programming framework. The approach establishes a PTAS for unit disks and disk-like regions under α-fat E or convexity assumptions, though the existence of a PTAS for general α-fat regions remains open due to unresolved connectivity and structural issues.

ABSTRACT

The Euclidean TSP with neighborhoods (TSPN) is the following problem: Given a set R of k regions, find a shortest tour that visits at least one point from each region. We study the special cases of disjoint, connected, alpha-fat regions (i.e., every region P contains a disk of diameter diam(P)/alpha) and disjoint unit disks. For the latter, Dumitrescu and Mitchell proposed an algorithm based on Mitchell's guillotine subdivision approach for the Euclidean TSP and claimed it to be a PTAS. However, their proof contains a severe gap, which we will close in the following. Bodlaender et al. remark that their techniques for the minimum corridor connection problem carry over to the TSPN and yield an alternative PTAS for this problem. For disjoint connected alpha-fat regions of varying size, Mitchell proposed a slightly different PTAS candidate. We will expose several further problems and gaps in this approach. Some of them we can close, but overall, for alpha-fat regions, the existence of a PTAS for the TSPN remains open.

Motivation & Objective

• To identify and correct fundamental flaws in Mitchell’s claimed PTAS for TSP with neighborhoods among disjoint unit disks.
• To establish a rigorous PTAS for TSPN with disjoint unit disks using a modified guillotine subdivision and grid-rounding technique.
• To investigate whether the guillotine approach can be extended to general α-fat or α-fat E regions, particularly addressing connectivity and structural integrity issues.
• To clarify the limitations of existing approaches and determine the current state of PTAS existence for TSPN under various fatness and geometric constraints.

Proposed method

• Reformulates the TSPN as a dynamic programming problem over recursively subdivided windows using axis-aligned cuts, enforcing an (m,M)-guillotine property for bounded complexity.
• Applies a grid-rounding scheme with δ = (2⌈k/ε⌉)⁻² to ensure polynomial-size state space while preserving tour length within (1+O(ε)) of optimal.
• Introduces a charging scheme to bound the increase in tour length due to grid rounding and edge corrections, ensuring total length remains within (1+O(ε)) of optimal.
• Uses a recursive construction starting from an optimal tour on a perturbed grid, inserting edges to maintain (m+9,M+24)-guillotine structure and connectivity.
• Employs a modified notion of α-fatness (α-fat E) to ensure bounded perimeter-to-diameter ratios, enabling length and connectivity guarantees.
• Validates the approach through lemmas on edge correction, parity adjustment, and subwindow compatibility, ensuring Eulerian and connected edge sets.

Experimental results

Research questions

• RQ1 Can Mitchell’s original PTAS claim for TSPN with disjoint unit disks be formally validated, or are there irreparable flaws in the argument?
• RQ2 What modifications to the guillotine subdivision and grid-rounding process are required to ensure both (1+O(ε))-approximation and connectivity?
• RQ3 To what extent can the PTAS framework be extended to general α-fat regions, and what geometric assumptions are necessary?
• RQ4 Why does the original approach fail to preserve connectivity in the presence of external regions, and how can this be repaired?
• RQ5 Is there a PTAS for TSPN with disjoint, connected α-fat regions of varying size, or does the problem remain open?

Key findings

• The original PTAS claim by Dumitrescu and Mitchell for disjoint unit disks contains a severe gap, which is fully closed in this work.
• A (1+O(ε))-approximation algorithm for TSPN with disjoint unit disks is established using a refined guillotine subdivision and grid-rounding method.
• An edge set E exists that satisfies the (m+9,M+24)-guillotine property, is Eulerian and connected, and has total length (1+O(ε))L*, where L* is the optimal tour length.
• The charging scheme successfully bounds the length increase from grid rounding and edge corrections, ensuring the approximation ratio is preserved.
• For α-fat E regions or convex regions, the framework can be adapted to maintain the (1+O(ε))-approximation, but connectivity remains a challenge for regions of varying size.
• The existence of a PTAS for general α-fat regions (in Mitchell’s sense) remains open due to unresolved issues in preserving connectivity and structural integrity in recursive subdivisions.