Design and Simulation of the Adaptive Continuous Entanglement Generation Protocol

TL;DR

The paper tackles reducing time-to-serve ($TTS$) for distributing end-to-end entangled pairs in quantum networks by proposing Adaptive Continuous entanglement Generation Protocol (ACP), which pre-generates EPs in the background and adaptively selects neighbors to optimize latency. ACP is implemented as a set of extensions to the SeQUeNCe quantum network simulator, including a single-heralded entanglement generation protocol aware of pre-generated EPs, a resource reservation mechanism, and a purification-aware policy. Across simulations on three network scales (2-, 20-, and 200-node topologies), ACP achieves significant $TTS$ reductions (roughly 57% to 94%) and fidelity improvements (up to about 0.05), validating the approach and its adaptability to changing traffic. The work highlights the practical impact of background entanglement provisioning combined with purification and adaptive routing aids for scalable quantum networks.

Abstract

Generating and distributing remote entangled pairs (EPs) is a primary function of quantum networks, as entanglement is the fundamental resource for key quantum network applications. A critical performance metric for quantum networks is the time-to-serve (TTS) for users' EP requests, which is the time to distribute EPs between the requested nodes. Minimizing the TTS is essential given the limited qubit coherence time. In this paper, we study the Adaptive Continuous entanglement generation Protocol (ACP), which enables quantum network nodes to continuously generate EPs with their neighbors, while adaptively selecting the neighbors to optimize TTS. Meanwhile, entanglement purification is used to mitigate decoherence in pre-generated EPs prior to the arrival of user requests. We extend the SeQUeNCe simulator to fully implement ACP and conduct extensive simulations across various network scales. Our results show that ACP reduces TTS by up to 94% and increases entanglement fidelity by up to 0.05.

Figures (10)

• Figure 1: Toy example of a quantum network, request, and time to serve (TTS). A request arrives at node 0 (initiator), asking node 0 to generate 1 EP with node 4 (responder) within the [1, 2] seconds time range, and requires the fidelity greater than 0.7. If the EP is generated at time $t = 1.2$ seconds, then the TTS of this request is 0.2 seconds.
• Figure 2: Toy example of the problem. The ACP runs on Node-A, which has five neighbors in total. ACP is allowed to use a maximum of four quantum memories for continuous generation of link EPs. Node-A in (a) has to decide what neighbors to generate EP with. In (b), the user requests result in paths (computed from the entanglement routing algorithm) that frequently include the segments [2, A, 3] and [3, A, 5]. Thus, Node-A should select neighbor Node-3 the most, followed by Node-2 and Node-5.
• Figure 3: The two FSMs of ACP. (a) depicts the FSM at Node-A, while (b) shows the FSM at Node-A's neighbors. Note that ACP is a symmetric peer-to-peer protocol, so the FSM in (a) is also running on Node-A's neighbor, and the FSM in (b) is also running on Node-A. This figures depicts "one connection". We achieved multiple connections by having multiple $S_2$ states in the Node-A FSM.
• Figure 4: (a) Node-$A$'s probability table. (b) Node-$i$ receives a request to generate EPs with Node-$r$. After the request is served, Node-$i$ and Node-$r$ will update their probability tables. Meanwhile, Node-$i$ sends a message containing the path to the intermediate nodes. Intermediate nodes update their probability tables after receiving the message.
• Figure 5: SeQUeNCe architecture has six modules. Each module has several components (not all components are shown). Subsection \ref{['subsec:quantum_manager']} to \ref{['subsec:reservation']} discuss the extensions in the eight highlighted components in yellow.