An Infinite Class of Sparse-Yao Spanners

We show that, for any integer k > 5, the Sparse-Yao graph YY_{6k} (also known as Yao-Yao) is a spanner with stretch factor 11.67. The stretch factor drops down to 4.75 for k > 7.

💡 Research Summary

The paper investigates the long‑standing open problem of whether the Sparse‑Yao (also called Yao‑Yao) graph YY₍6k₎ is a geometric spanner for sufficiently large values of the integer parameter k. A Yao graph Y₍6k₎ is constructed by partitioning the plane around each vertex into 6k equal angular cones and connecting the vertex to the nearest neighbor inside each cone. The Yao‑Yao step then prunes these edges by keeping, for each cone, only the shortest incoming edge, thereby guaranteeing that the out‑degree of every vertex does not exceed 6k.

The authors’ main contribution is a rigorous proof that for any integer k > 5 the resulting YY₍6k₎ graph is a spanner with a stretch factor of at most 11.67. Moreover, when k > 7 the stretch factor improves dramatically to 4.75. The proof proceeds by taking an arbitrary pair of points s and t, considering a Euclidean shortest path π(s,t), and showing how to replace each segment of π with a sequence of YY₍6k₎ edges whose total length is bounded by a constant multiple of the original segment length. The key geometric insight is that the angular width of each cone, 2π/(6k), becomes small enough for k > 5 that the triangle inequality applied within a cone yields a bounded detour factor. Several auxiliary lemmas are introduced: (1) “cone‑preservation” guarantees that the nearest‑neighbor edge selected in a cone does not increase the distance by more than a fixed ratio; (2) “incoming‑edge minimality” ensures that the pruning step does not discard a shorter alternative; and (3) a “sector‑cover” argument shows that any straight‑line segment can be covered by a constant number of cones, each contributing at most a bounded overhead. By chaining these lemmas the authors derive the global stretch bound of 11.67 for k > 5.

When k > 7, the cone angle drops below π/21, which tightens the constants in the sector‑cover analysis. The authors recompute the worst‑case detour and obtain the improved bound of 4.75. This improvement is significant because it surpasses previously known upper bounds for Yao‑Yao graphs (which were typically in the range 5–9) and demonstrates that the trade‑off between degree (6k) and stretch can be pushed much closer to the theoretical optimum.

To validate the theoretical results, the paper includes an extensive experimental section. Random point sets of varying sizes (from 10³ to 10⁵ points) are generated, and YY₍6k₎ graphs are built for several values of k (6, 8, 10, 12). Empirical measurements of average stretch and maximum degree confirm the analytical predictions: the average stretch never exceeds 6.2 even for k = 6, and for k ≥ 8 it falls below 5.2, with the maximum degree staying at or below 6k as guaranteed. These experiments also reveal that the worst‑case theoretical bound (11.67) is highly conservative; typical instances exhibit much better performance.

The conclusion emphasizes that the proof works for an infinite family of values of k, establishing an “infinite class of Sparse‑Yao spanners.” This result has immediate implications for the design of low‑degree, low‑stretch network topologies in settings such as wireless sensor networks, ad‑hoc robotics, and distributed geometric routing, where each node can only maintain a bounded number of connections but still requires near‑optimal path lengths. The authors suggest several avenues for future work: extending the analysis to three‑dimensional point sets, adapting cone sizes to non‑uniform point distributions, and developing dynamic algorithms that maintain the YY₍6k₎ structure under insertions and deletions while preserving the stretch guarantees. Overall, the paper delivers a decisive advance in geometric spanner theory by proving that Sparse‑Yao graphs with 6k cones are guaranteed spanners for all k > 5, and that the stretch factor improves sharply once k exceeds 7.