HAQA: a hardware-guided and fidelity-aware strategy for efficient qubit mapping optimization

Abstract Quantum compilation bridges the gap between abstract quantum algorithms and physical hardware execution. However, on Noisy Intermediate-Scale Quantum (NISQ) devices, particularly within superconducting architectures, mapping logical circuits to physical qubits remains a fundamental bottleneck due to rigid connectivity constraints and heterogeneous noise profiles. While exact mapping solvers theoretically ensure optimality, they face severe scalability issues on large-scale chips. The search space for global optimization becomes intractable as the system size increases. This scalability barrier largely arises because conventional methods often treat the hardware coupling graph as a homogeneous entity. By performing global searches with weak awareness of physical fidelity distributions, they fail to exploit the hardware’s inherent structural advantages. To address this, we propose HAQA, a hardware-guided strategy that shifts the mapping paradigm from a blind global search to an adaptive, fidelity-aware regional optimization. Specifically, HAQA identifies optimal topological subgraphs through a community-based recursive fusion mechanism. This approach achieves a dual benefit: it reduces solving complexity from exponential to polynomial levels by effectively pruning the search space, while simultaneously enhancing execution fidelity by prioritizing high-quality physical qubits. Experimental evaluations on IBM Eagle and Heron processors demonstrate significant efficiency gains. HAQA accelerates state-of-the-art SMT solvers (QSynth-v2 and TB-OLSQ2) by factors of 632.76× and 286.87×, respectively, while improving fidelity by up to 238.28%. These results suggest that integrating fine-grained physical awareness provides an effective pathway to balance computational complexity and execution quality in quantum compilation, positioning HAQA as a plug-and-play enhancement compatible with various solvers.