arXiv2025
The presence of obstacles has a major impact on distance computation, motion planning, and visibility. While these problems are well studied in the plane, our understanding in three and higher dimensions is still limited. We investigate how different obstacle properties affect the induced geodesic metric in three-dimensional space. A finite metric space is said to be approximately realizable by a collection of obstacles if, for any $\varepsilon>0$, it can be embedded into the free space around the obstacles with geodesic distance and worst-case distortion $1+\varepsilon$. We focus on three key properties-convexity, disjointness, and fatness-and analyze how omitting each of them influences realizability. Our main result shows that if fatness is dropped, then every finite metric space can be realized with distortion $1+\varepsilon$ using convex, pairwise disjoint obstacles in $\mathbb{R}^3$, even if all obstacles are congruent equilateral triangles. Moreover, if we enforce fatness but drop convexity or disjointness, the same realizability still holds. Our results have important implications on the approximability of TSP with Obstacles, a natural variant of TSP introduced recently by Alkema et al. (ESA 2022). Specifically, we use the recent results of Banerjee et al. on TSP in doubling spaces (FOCS 2024) and of Chew et al. on distances among obstacles (Information Processing Letters 2002) to show that TSP with Obstacles admits a PTAS if the obstacles are convex, fat, and pairwise disjoint. If any of these three properties is dropped, then our results, combined with the APX-hardness of Metric TSP, demonstrate that TSP with Obstacles is APX-hard.