Learning-Optimized Qubit Mapping and Reuse to Minimize Inter-Core Communication in Modular Quantum Architectures

TL;DR

The paper tackles the bottleneck of inter-core communication in modular quantum architectures by introducing QARMA, a transformer-encoder/GNN-based DRL framework for qubit allocation, and QARMA-R, which adds dynamic qubit reuse via mid-circuit measurement. The authors design a comprehensive environment that identifies reuse opportunities, a policy model with initial/context/dynamic embeddings and a pointer mechanism, and a transformer-based encoder-decoder to capture both local and global circuit structure. Empirical results show dramatic reductions in inter-core transfers—up to 100% (average ~86%) with reuse, and significant improvements even without reuse for larger circuits—compared to QUBO-based mappers and Qiskit-O3, with favorable computational efficiency. A hardware-calibrated fidelity analysis demonstrates that the benefits of qubit reuse far outweigh the modest depth increases, highlighting the practical impact for enabling larger, higher-fidelity quantum computations on resource-constrained modular systems.

Abstract

Modular quantum architectures have emerged as a promising approach for scaling quantum computing systems by connecting multiple Quantum Processing Units (QPUs). However, this approach introduces significant challenges due to costly inter-core operations between chips and quantum state transfers, which contribute to noise and quantum decoherence. This paper presents QARMA, a novel Qubit mapping using Attention-based deep Reinforcement learning (DRL) for Modular quantum Architectures, along with its extension QARMA-R that incorporates dynamic qubit reuse capabilities. Our approach combines an attention-based mechanism with Graph Neural Networks (GNN) to learn optimal qubit allocation, routing, and reuse strategies that minimize inter-core communications. We introduce two key innovations: (1) a transformer-based encoder that captures both the global circuit structure and local qubit interactions and (2) a dynamic qubit reuse compilation mechanism that leverages mid-circuit measurement and reset operations to reduce inter-operation and qubit requirements. Our experimental results show significant improvements over state-of-the-art approaches. Compared to highly-optimized Qiskit with modular architecture configuration, QARMA-R reduces inter-core communications by up to 100% (on average 86%), while QARMA maintains 15-40% improvement for larger circuits without reuse. Against traditional modular qubit mapping, our approach achieves 97-100% reduction in inter-core operation. The proposed methods advance quantum circuit compilation techniques and enable the execution of more extensive quantum algorithms on resource-constrained modular quantum systems, contributing to the growing body of research on scalable quantum computing architectures.

Figures (6)

• Figure 1: Quantum circuit compilation process for the Bernstein-Vazirani (BV) algorithm bv_alg quantum circuit. a) Original logical circuit with five qubits, showing Hadamard (H) gates and a CNOT operation. b) Physical architecture with limited connectivity between qubits. c) Compiled circuit with inserted SWAP gates (red) to accommodate hardware connectivity constraints.
• Figure 2: Using Dynamic Circuit Support for the Bernstein-Vazirani (BV) algorithm bv_alg quantum circuit to reduce qubit usage: a) original five-qubit circuit; b) qubit 0 reused once, reducing to four qubits; c) qubit 0 reused three times, requiring only two qubits. Vertical double lines indicate measurement/reset operations.
• Figure 3: Modular quantum architecture and circuit execution: a) A multi-core quantum architecture with two cores, each containing four fully-connected physical qubits, linked by inter-core connections; b) A sliced quantum circuit where each vertical region represents a time slice containing gates that can be executed in parallel; c) Qubit allocation and state transfers across time slices, showing how logical qubits (numbered 0-4) are mapped to physical cores ($c_0$ and $c_1$) at different time steps ($T_0$ through $T_6$). Arrows indicate quantum state transfers between cores, which are more costly than intra-core operations.
• Figure 4: Overall of our approach for qubit allocation with reuse in modular quantum architectures. a) Input quantum circuit with multiple qubits and gates; b) environment that performs dependency analysis to identify qubit reuse opportunities and generate action masks; c) an attention-based policy model with encoder-decoder architecture that processes the circuit's interaction graph and produces allocation decisions; d) final allocation solution showing quantum state transfers between cores across time steps. The numbered arrows (1-7) indicate the workflow sequence through our approach.
• Figure 5: Comparison of inter-operation qubit communication or inter-core connection counts using Qiskit. The baseline uses Qiskit 1.4.0, with the highest optimization level (Level 3). Bars show the absolute number of inter-operation counts. Our method was evaluated in two settings: QARMA, with qubit reuse disabled, and QARMA-R, with reuse enabled. The inset highlights results for smaller circuits.