CARL : crosstalk-aware quantum compilation based on reinforcement learning

Abstract

Quantum Computing (QC) is promising for solving complex problems that are intractable for classical computers. However, QC on current quantum processors faces a critical challenge in achieving high execution fidelity due to decoherence and accumulated gate errors, limiting its practical applicability. Quantum circuit compilation (QCC), which transforms highlevel quantum circuits into hardware-compliant implementations, directly affects execution fidelity. Existing work either neglects hardware constraints such as qubit connectivity or does not explicitly optimize for execution fidelity, which can significantly degrade circuit fidelity. This paper investigates the problem of circuit compilation, which jointly considers qubit connectivity constraints and optimizes execution fidelity explicitly. To address the problem, this paper presents a reinforcement learning–based QCC approach named CARL. In particular, CARL unifies mapping and routing into a single crosstalk-aware optimization pipeline to mitigate hardware noise. Extensive evaluations on real-world benchmark quantum circuits demonstrate that the proposed approach achieves significantly higher circuit fidelity than state-of-the-art baselines.