QAISim: A Toolkit for Modeling and Simulation of AI in Quantum Cloud Computing Environments

TL;DR

The paper addresses the challenge of efficiently managing quantum hardware resources in cloud environments for IoT workloads. It introduces QAISim, a Python-based toolkit that uses parameterized quantum circuits to implement reinforcement learning policies (Policy Gradient and Deep Q-Learning) for quantum resource management, with a greedy baseline for comparison. Through simulations using Cirq, TFQ, and Gymnasium, QAISim demonstrates competitive performance relative to classical DRL while requiring far fewer trainable parameters, highlighting the potential for quantum ML to optimize QCaaS. The work also outlines future directions, including multi-objective optimization, hardware experiments, and AIoT integration, to advance practical QAISim deployment in real quantum-cloud systems.

Abstract

Quantum computing offers new ways to explore the theory of computation via the laws of quantum mechanics. Due to the rising demand for quantum computing resources, there is growing interest in developing cloud-based quantum resource sharing platforms that enable researchers to test and execute their algorithms on real quantum hardware. These cloud-based systems face a fundamental challenge in efficiently allocating quantum hardware resources to fulfill the growing computational demand of modern Internet of Things (IoT) applications. So far, attempts have been made in order to make efficient resource allocation, ranging from heuristic-based solutions to machine learning. In this work, we employ quantum reinforcement learning based on parameterized quantum circuits to address the resource allocation problem to support large IoT networks. We propose a python-based toolkit called QAISim for the simulation and modeling of Quantum Artificial Intelligence (QAI) models for designing resource management policies in quantum cloud environments. We have simulated policy gradient and Deep Q-Learning algorithms for reinforcement learning. QAISim exhibits a substantial reduction in model complexity compared to its classical counterparts with fewer trainable variables.

Figures (8)

• Figure 1: Top-Level abstraction of a quantum cloud computing environment and a sample task execution workflow.
• Figure 2: Important quantum gates (unitary operators) including hadamard, control and rotation gates.
• Figure 3: A 3-qubit data re-uploading parametrized quantum circuit with $U_{\text{var}}(\phi_{i})$ and $U_{\text{enc}}(s, \lambda_{i})$ layers with variational parameters $\phi_{i}$ (rotational angles) and $\lambda_{i}$ (scaling parameters) and input parameters $s$. We use CZ operators to perform the entangling operations in $U_{\text{var}}(\phi_{i})$.
• Figure 4: QAISim Architecture.
• Figure 5: Fundamental Classes of QAISim.