RADAR-Q: Resource-Aware Distributed Asynchronous Routing for Entanglement Distribution in Multi-Tenant Quantum Networks

Abstract: Scalable quantum networks must support concurrent entanglement requests, yet existing routing protocols fail when users compete for shared repeater resources, wasting fragile quantum states. This paper presents RADAR-Q, a resource-aware decentralized routing protocol embedding real-time resource contention into path selection. Unlike prior designs requiring global coordination or central anchors, RADAR-Q makes intelligent local decisions balancing path length and fidelity, instantaneous quantum memory availability, and intermediate Bell-State Measurement (BSM) operations. By identifying the Nearest Common Ancestor (NCA) within a DODAG hierarchy, RADAR-Q localizes entanglement swapping close to communicating users - avoiding unnecessary central detours and reducing BSM chain length and decoherence exposure. We evaluate RADAR-Q on grid and random topologies against synchronous and root-centric asynchronous baselines. Results show RADAR-Q achieves aggregate throughputs 2.5x and 7.6x higher than synchronized and root-centric designs, respectively. While baselines suffer catastrophic fidelity collapse below the 0.5 threshold under high load, RADAR-Q consistently maintains end-to-end fidelity above 0.76, ensuring pairs remain usable. Furthermore, RADAR-Q exhibits near-perfect fairness (Jain's Fairness Index 96-98%) and retains over 50% of its ideal throughput under stringent 1.0 ms coherence times. These findings establish contention-aware decentralized routing as a scalable foundation for multi-tenant quantum networks.

• The paper introduces RADAR-Q, a distributed routing protocol that combines DODAG-based hierarchy, nearest-common-ancestor paths, link fidelity, and memory availability to arbitrate quantum-network contention locally without global synchronization.
• RADAR-Q delivers up to 7.6× higher throughput than asynchronous root-centric routing, maintains approximately 0.76–0.77 fidelity at 10 concurrent requests, and achieves Jain fairness scores of 0.96–0.98 in simulated grid and random networks.
• The protocol remains effective with 1.0 ms coherence times, but its results require validation under realistic control-plane delays, parameter sensitivity, photonic BSM success rates, differentiated service policies, and adversarial multi-tenant conditions.

Motivation and problem context

Multi-tenant quantum networks require a routing layer that can serve many concurrent entanglement requests over shared repeater infrastructure. The authors' prior protocols, AER and MAER, demonstrated efficient asynchronous routing for isolated sessions but explicitly lack mechanisms for distributed resource arbitration: when multiple source–destination (S–D) pairs independently select overlapping paths through a common repeater, each may reserve qubits for Bell-State Measurement (BSM) without coordination, and because quantum memory is finite and non-bufferable under decoherence, all competing requests can block indefinitely. Synchronous protocols such as slot-based approaches avoid this by global coordination, but their scheduling overhead grows with network size and concurrency, and they discard all unused entanglement at slot boundaries. Shortest-path asynchronous routing over instantaneous link graphs ignores real-time memory availability entirely.

RADAR-Q's central claim is that contention-awareness, rather than asynchronicity alone, is the decisive factor for scalable multi-tenant routing. The protocol embeds resource competition directly into path selection using only local state, eliminating both global coordination and root-centric funneling.

Protocol design

RADAR-Q builds on a DODAG hierarchy borrowed from RPL, maintained via DIO, DIS, and DAO control messages extended with quantum-specific parameters: per-node average link fidelity $F_v$ and memory utilization $Q_v/M_v$. Each node computes a composite rank

$$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$

where the fractional term is bounded in $[0,1)$. This bound guarantees strict rank monotonicity along downward paths ($\text{rank}(v) < \text{rank}(\text{parent}(v)) + 1$), providing a loop-free guarantee while permitting quantum-aware tie-breaking within equal hop distance.

Path selection is NCA-centric: for any S–D pair, entanglement swapping is localized at the Nearest Common Ancestor—the deepest common ancestor in the DODAG—yielding two upward segments whose total hop count $k'$ satisfies $k' \leq k$, where $k$ is the root-routed path length. This directly reduces BSM depth, which governs both aggregate success probability ($\propto q^k$) and cumulative decoherence exposure. Requests are processed in descending order of NCA depth ("locality-first"), and among candidate paths the protocol maximizes

$$\frac{d_{\text{hop}}(\text{nca}(s,d)) + 1}{1 + \max_{e \in p}(1 - \text{avail}(e))},$$

rewarding locality in the numerator and penalizing congestion via worst-link availability in the denominator. Paths containing saturated links are discarded; on BSM failure or decoherence, lightweight notifications trigger localized DODAG updates rather than global reconvergence. Ancestor discovery is fully distributed: nodes broadcast ancestor lists upward via parent pointers, and the first intersection yields the NCA.

The design assumes heralded link generation with reported success probability $Q_v/M_v$0 and fidelity $Q_v/M_v$1, and is agnostic to the specific physical scheme (Emitter–Scatter, Emitter–Emitter, or Scatter–Scatter).

Evaluation methodology

RADAR-Q is compared against two architectural extremes: Synch-NCA, a globally synchronized NCA-based protocol treated as a fidelity upper bound, and Asynch-Root, a standard distributed root-centric baseline without contention awareness. Two topologies are used—a $Q_v/M_v$2 grid and a random graph ($Q_v/M_v$3, average degree 4)—chosen to isolate contention-aware resource arbitration without presupposing a physical architecture. Parameters target near-term matter-qubit platforms: link generation probability $Q_v/M_v$4, BSM success probability $Q_v/M_v$5 (consistent with NV centers and trapped ions supporting deterministic BSMs), and initial fidelity $Q_v/M_v$6. Fidelity degradation combines exponential memory decay $Q_v/M_v$7 with multiplicative BSM infidelity scaling as $Q_v/M_v$8. Scalability tests assume $Q_v/M_v$9 to isolate protocol logic; robustness tests vary $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$0 from 1.0 ms upward.

A caveat worth noting: the chosen $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$1 reflects matter-qubit repeaters, whereas linear-optical photonic BSMs are fundamentally limited to $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$2; results should be read against that hardware regime.

Throughput scalability

Under idealized coherence, RADAR-Q exhibits near-linear throughput growth. At $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$3 concurrent requests it achieves approximately 3.8 pairs/sec on the grid—2.5× higher than Synch-NCA and 7.6× higher than Asynch-Root—and approximately 4.5 pairs/sec on the random topology. The gap widens with concurrency, from roughly 2× at $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$4 to over 7× at $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$5, indicating that the advantage stems from load distribution across parallel NCA-localized paths rather than constant-factor overhead reduction. Notably, RADAR-Q outperforms even the globally coordinated Synch-NCA, which suffers scheduling overhead and lacks proactive handling of overlapping paths. This implies that decentralized contention-aware selection can dominate centralized scheduling in throughput once concurrency is high enough that coordination cost itself becomes a bottleneck.

Fidelity–throughput trade-off

Synch-NCA retains the highest end-to-end fidelity (~0.95), as expected from its idealized coordination. RADAR-Q stabilizes around 0.76–0.77 at $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$6 in both topologies—a modest penalty the authors frame as the necessary price of removing global synchronization. In contrast, Asynch-Root collapses below the 0.5 distillation threshold (0.48 on the grid, 0.36 on the random graph), rendering generated pairs physically unusable for teleportation or error correction. The mechanism behind RADAR-Q's stability is structural: by minimizing BSM depth $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$7, it avoids the exponential success-probability penalty $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$8 and the extended storage durations that induce retry-driven decoherence. The practical implication is that every pair RADAR-Q produces remains distillable and usable downstream, whereas naive asynchronous operation produces largely invalid entanglement under load.

Fairness

Using Jain's Fairness Index, RADAR-Q maintains 0.96–0.98 across both topologies at all tested concurrencies, indicating near-uniform service distribution. Asynch-Root degrades severely—to 0.39 on the grid and 0.24 on the random topology at $$\text{rank}(v) = d_{\text{hop}}(v) + \frac{\alpha(1-F_v) + \beta(Q_v/M_v)}{\alpha+\beta},$$9—because nodes near the root monopolize links and starve peripheral users. Embedding real-time link availability into the metric steers requests toward underutilized NCAs, preventing resource monopolization without explicit quota mechanisms. For multi-tenant deployments where service predictability is contractual, this fairness property is arguably as consequential as raw throughput.

Robustness to finite coherence time

Varying $[0,1)$0 from 1.0 ms to infinity, RADAR-Q preserves linear throughput growth across all regimes. At $[0,1)$1 ms with $[0,1)$2 on the random topology, it delivers approximately 2.1 pairs/sec—over 50% of its idealized throughput. Because NCA-centric paths cap qubit storage duration, finite coherence imposes an approximately constant multiplicative penalty rather than triggering the load-dependent collapse seen in root-centric designs. This decoupling of scalability from hardware volatility is what makes the protocol plausibly deployable on NISQ-era platforms where coherence is the binding constraint.

Limitations and open questions

Several constraints qualify these results. First, the evaluation is simulation-only, with idealized classical signaling assumed; DODAG maintenance latency, message loss, and convergence time under churn are not modeled, though RPL-style hierarchies are known to be sensitive to these in practice. Second, the rank weights $[0,1)$3 and $[0,1)$4 are tunable but no sensitivity analysis or tuning methodology is provided, leaving their selection open. Third, the contention metric penalizes only the single worst link along a path ($[0,1)$5), which may misestimate aggregate congestion on longer candidate paths. Fourth, the best-effort model assigns uniform weight to all requests; differentiated service classes, priority scheduling, and admission control remain unaddressed. Fifth, the parameter regime ($[0,1)$6, $[0,1)$7) excludes photonic-interconnect scenarios with $[0,1)$8, and performance there is unverified. Finally, the paper does not evaluate security or adversarial behavior in multi-tenant settings—an important question for shared quantum infrastructure.

Conclusion

RADAR-Q demonstrates that embedding real-time resource contention into locally computable path metrics resolves the multi-tenant bottleneck that defeats both synchronous and root-centric asynchronous designs. Its combination of up to 7.6× throughput gains, sustained fidelity above the 0.5 distillation threshold (~0.76), near-perfect fairness (>0.96), and robustness down to 1.0 ms coherence times establishes contention-aware decentralized routing as a credible foundation for shared quantum networks. The principal open questions concern validation under realistic classical-control impairments, sensitivity to the rank-weight hyperparameters, and extension beyond the uniform-request, matter-qubit regime evaluated here.