Quantum computing promises to execute some tasks exponentially faster than classical computers. Quantum compilation, which transforms algorithms into executable quantum circuits, involves solving the initial mapping problem, crucial for optimizing qubit assignment and minimizing gate error rates. This study explores Deep Reinforcement Learning (DRL) for initial mapping across various qubit topologies, considering fixed gate error rates. Previous DRL approaches have succeeded but didn’t account for fixed error rates, used only one algorithm (PPO), and focused on a single topology with 20 qubits. The trial-and-error nature of Reinforcement Learning makes it ideal for initial mapping. DRL agents, using multiple policy gradient algorithms (A2C, PPO with and without action masking, and TRPO), compute high-quality mappings for small- and medium-scale quantum architectures. While effective, their efficiency decreases with larger systems, necessitating further optimization. Fine-tuning hyperparameters and action masking prevent illegal actions and enhance accuracy. Although currently not surpassing tools like Qiskit or achieving scalability for larger systems, this study highlights DRL’s potential for initial mapping in quantum computing, encouraging further innovation and refinement.

Compiling a quantum circuit for specific quantum hardware is a challenging task. Moreover, current quantum computers have severe hardware limitations. To make the most use of the limited resources, the compilation process should be optimized. To improve currents methods, Reinforcement Learning (RL), a technique in which an agent interacts with an environment to learn complex policies to attain a specific goal, can be used. In this work, we present qgym, a software framework derived from the OpenAI gym, together with environments that are specifically tailored towards quantum compilation. The goal of qgym is to connect the research fields of Artificial Intelligence (AI) with quantum compilation by abstracting parts of the process that are irrelevant to either domain. It can be used to train and benchmark RL agents and algorithms in highly customizable environments.