arXiv2026
The stretch factor of a triangulation is the maximum, over all pairs of its vertices, of the ratio between their shortest-path distance in the triangulation and their Euclidean distance. Determining the worst-case stretch factor of planar Delaunay triangulations is a long-standing open problem in computational geometry; the best known lower and upper bounds have been 1.5932, due to Xia and Zhang, and 1.998, due to Xia. Xia's induction over the circumdisks of the triangles crossed by the segment between two vertices compares shortest paths along their circles with polylines whose lengths do not split into contributions of single disks, so it has to be completed by a global estimate that loses a constant factor. In this paper, we propose progress accounting: the chords shared by consecutive circumdisks cut the segment into parts, one for each disk, and we call the length of the part of a disk its progress; the progresses add up exactly to the length of the segment. The length of a shortest path along the circles also splits into local terms, one for each disk, which are computed from the shortest distances to the two ends of each shared chord; an upper bound U then follows from a potential function whose decrease pays, at every disk, for its term minus U times its progress. We show that such a potential exists whenever a function of four variables satisfies five explicit conditions. We then construct such a function, which satisfies one of the conditions by construction, and verify the other four by a computer-assisted proof in exact rational and interval arithmetic. As a result, every Delaunay triangulation of a finite planar point set has stretch factor at most 79742997/50000000 < 1.59486, which reduces the gap between the best known upper and lower bounds from 0.405 to less than 0.0017.