Throughput Scaling in Random Wireless Networks: A Non-Hierarchical Multipath Routing Strategy

Franceschetti et al. have recently shown that per-node throughput in an extended, ad hoc wireless network with $\Theta(n)$ randomly distributed nodes and multihop routing can be increased from the $\Omega({1 \over \sqrt{n} \log n})$ scaling demonstrated in the seminal paper of Gupta and Kumar to $\Omega({1 \over \sqrt{n}})$. The goal of the present paper is to understand the dependence of this interesting result on the principal new features it introduced relative to Gupta-Kumar: (1) a capacity-based formula for link transmission bit-rates in terms of received signal-to-interference-and-noise ratio (SINR); (2) hierarchical routing from sources to destinations through a system of communal highways; and (3) cell-based routes constructed by percolation. The conclusion of the present paper is that the improved throughput scaling is principally due to the percolation-based routing, which enables shorter hops and, consequently, less interference. This is established by showing that throughput $\Omega({1 \over \sqrt{n}})$ can be attained by a system that does not employ highways, but instead uses percolation to establish, for each source-destination pair, a set of $\Theta(\log n)$ routes within a narrow routing corridor running from source to destination. As a result, highways are not essential. In addition, it is shown that throughput $\Omega({1 \over \sqrt{n}})$ can be attained with the original threshold transmission bit-rate model, provided that node transmission powers are permitted to grow with $n$. Thus, the benefit of the capacity bit-rate model is simply to permit the power to remain bounded, even as the network expands.

💡 Research Summary

The paper revisits the fundamental scaling law for per‑node throughput in large, extended ad‑hoc wireless networks. Gupta and Kumar’s seminal result showed that, under a simple protocol‑model where each link succeeds if the received signal‑to‑interference‑plus‑noise ratio (SINR) exceeds a fixed threshold, the achievable throughput per node scales as Ω(1/(√n log n)) when n nodes are uniformly scattered over a unit‑density area and multihop routing is used. Franceschetti et al. later broke the logarithmic penalty and demonstrated a scaling of Ω(1/√n). Their improvement relied on three new ingredients: (1) a Shannon‑capacity based link rate formula that directly ties the transmission bit‑rate to the actual SINR, (2) a hierarchical routing scheme that routes traffic through a set of “highways” (long, dedicated backbone paths), and (3) a percolation‑based construction of cell‑level routes that yields many short hops.

The present work asks a natural question: which of these three ingredients is truly responsible for the better scaling? To answer this, the authors strip away the highway infrastructure entirely and retain only the percolation‑based routing idea. They define, for each source–destination pair, a narrow rectangular “routing corridor” that connects the two nodes. Classical bond‑percolation theory guarantees that, with high probability, Θ(log n) node‑disjoint paths exist inside this corridor, each consisting of hops of length Θ(1/√n). Because the hops are short, the interference generated by simultaneous transmissions remains bounded, and each hop can achieve a constant SINR even when all nodes transmit at a fixed power. Consequently, the network can support Ω(1/√n) throughput without any highways and without increasing the transmit power as n grows.

The authors also examine the role of the capacity‑based rate model. They prove that the same Ω(1/√n) scaling can be achieved under the original threshold‑rate model provided that node transmit powers are allowed to grow with n (e.g., proportionally to √n). In this regime the SINR at each receiver stays above the fixed threshold, and the per‑hop capacity becomes effectively constant. Hence the capacity‑based model does not fundamentally change the scaling law; it merely permits the power to remain bounded while still achieving the optimal scaling.

Technical contributions include: (i) a rigorous probabilistic analysis showing that the routing corridor contains enough disjoint paths to support Θ(log n) simultaneous flows; (ii) a scheduling argument that demonstrates how to activate a constant fraction of these paths in each time slot while keeping aggregate interference O(1); (iii) a derivation of the required power scaling for the threshold model and a comparison with the capacity model.

The key insight is that the percolation‑based multipath routing, by creating many short hops, is the dominant factor that reduces interference and thus lifts the logarithmic penalty present in the Gupta‑Kumar analysis. Highways, while useful for simplifying routing decisions in practice, are not essential for achieving the Ω(1/√n) throughput scaling. This conclusion suggests that future designs of large‑scale wireless ad‑hoc networks should prioritize mechanisms that guarantee short, dense, and redundant paths—such as percolation‑inspired topology control—over hierarchical backbone constructions. Moreover, the paper clarifies that the apparent advantage of the Shannon‑capacity rate model is limited to keeping transmit power bounded; the same scaling can be obtained with the simpler threshold model if one is willing to let power increase with network size. Overall, the work isolates the true source of the scaling improvement and provides a clear roadmap for leveraging percolation theory in practical network protocol design.