ABSTRACT Quantum computing has entered the noisy intermediate‐scale quantum (NISQ) era, where limited qubit numbers, short coherence times, and high error rates pose significant challenges to reliable large‐scale execution. Efficient scheduling and resource allocation across heterogeneous quantum hardware are therefore crucial for maximizing system throughput, fidelity, and fairness. In this work, we propose a hardware‐aware reinforcement learning framework for quantum task scheduling in distributed NISQ systems. Our design explicitly models qubit‐level variability, including connectivity degree, coherence times, error rates, and throughput, while integrating task‐level constraints such as deadlines, priorities, and concurrency requirements. We formulate the scheduling problem as a constrained optimization task and instantiate it as a Markov decision process (MDP), enabling reinforcement learning agents to learn adaptive strategies. Specifically, we implement deep Q‐network (DQN) and proximal policy optimization (PPO) agents, and compare them against heuristic and random baselines. Experimental results demonstrate that PPO consistently outperforms DQN and heuristic methods, achieving higher task completion rates, fewer deadline violations, and more robust adaptation across different reward configurations. This work bridges quantum hardware modeling with reinforcement learning‐based scheduling, providing a practical pathway for resource optimization in distributed quantum computing environments.
ABSTRACT Quantum computing has entered the noisy intermediate‐scale quantum (NISQ) era, where limited qubit numbers, short coherence times, and high error rates pose significant challenges to reliable large‐scale execution. Efficient scheduling and resource allocation across heterogeneous quantum hardware are therefore crucial for maximizing system throughput, fidelity, and fairness. In this work, we propose a hardware‐aware reinforcement learning framework for quantum task scheduling in distributed NISQ systems. Our design explicitly models qubit‐level variability, including connectivity degree, coherence times, error rates, and throughput, while integrating task‐level constraints such as deadlines, priorities, and concurrency requirements. We formulate the scheduling problem as a constrained optimization task and instantiate it as a Markov decision process (MDP), enabling reinforcement learning agents to learn adaptive strategies. Specifically, we implement deep Q‐network (DQN) and proximal policy optimization (PPO) agents, and compare them against heuristic and random baselines. Experimental results demonstrate that PPO consistently outperforms DQN and heuristic methods, achieving higher task completion rates, fewer deadline violations, and more robust adaptation across different reward configurations. This work bridges quantum hardware modeling with reinforcement learning‐based scheduling, providing a practical pathway for resource optimization in distributed quantum computing environments.
This paper addresses the critical challenge of optimizing resource allocation within dynamic quantum networks (QNs). The inherent complexities of QNs, including fluctuating demands, evolving topologies, and the fragility of quantum states, necessitate adaptive resource management strategies. We propose a novel algorithm leveraging reinforcement learning (RL) to dynamically adjust resource allocation based on real-time network conditions. The algorithm learns an optimal policy by iteratively interacting with a simulated QN environment, maximizing the overall quantum channel throughput while minimizing the probability of quantum error. Specifically, the core mechanism involves formulating the resource allocation problem as a Markov Decision Process (MDP), where the state represents the network's current configuration, the action represents the allocation of quantum resources (e.g., qubits, entanglement), and the reward reflects the successful completion of quantum communication tasks. The RL agent utilizes a Q-learning algorithm to estimate the optimal Q-function, guiding the allocation process. We demonstrate, through simulation, that our algorithm significantly outperforms static allocation strategies in terms of throughput and reliability, even under highly dynamic network conditions. The key contributions of this work lie in the integration of RL with the unique characteristics of QNs and the development of a practical algorithm for adaptive resource management. Further research will explore the scalability of this approach and its application to larger, more complex QN topologies.
Quantum cloud computing enables remote access to quantum processors, yet the heterogeneity and noise of available quantum hardware create significant challenges for efficient resource orchestration. These issues complicate the optimisation of quantum task allocation and scheduling, as existing heuristic methods fall short in adapting to dynamic conditions or effectively balancing execution fidelity and time. Here, we propose QFOR, a Q uantum F idelity-aware O rchestration of tasks across heterogeneous quantum nodes in cloud-based environments using Deep R einforcement learning. We model the quantum task orchestration as a Markov Decision Process and employ the Proximal Policy Optimisation algorithm to learn adaptive scheduling policies, using IBM quantum processor calibration data for noise-aware performance estimation. Our configurable framework balances overall quantum task execution fidelity and time, enabling adaptation to different operational priorities. Extensive evaluation demonstrates that QFOR is adaptive and achieves significant performance with 29.5-84% improvements in relative fidelity performance over other deep reinforcement learning and heuristic baselines. Furthermore, it maintains comparable quantum execution times, contributing to cost-efficient use of quantum computation resources.
Abstract Quantum communication plays a key role in enabling distributed Noisy Intermediate-Scale Quantum (NISQ) systems; however, practical quantum networks are often affected by link fluctuations, limited quantum resources, and entanglement degradation. These factors make reliable routing and scheduling difficult, particularly in large-scale and dynamically changing environments. Classical methods such as Historical Weighted Fidelity Routing (HWFR) and Dynamic Multi-Priority Quantum Scheduling (DMPQS) generally perform well only when network conditions remain stable, and their performance degrades when entanglement success rates vary or relay nodes become congested. A major challenge therefore lies in operating under uncertain and time-varying fidelity conditions. Most existing approaches focus on either routing or scheduling in isolation and do not explicitly consider future fidelity variations, which can lead to inefficient resource usage and increased end-to-end latency. In this work, we propose a Deep Reinforcement Learning-based Quantum Communication Optimization Framework (DRL-QCOF) that jointly addresses routing and scheduling. The framework employs a Bi-LSTM-based Quantum Fidelity Prediction Network (QFPN) to estimate future link fidelity F_"pred" (t)from time-series features including link length, noise coefficient, entanglement success probability, and relay utilization. These predictions are then used by a Deep Reinforcement Learning-based Adaptive Quantum Scheduling Agent (DRAQS-Agent), implemented using an Actor–Critic architecture with a continuous reward formulation, to guide routing and scheduling decisions across distributed nodes. Simulation results indicate that the proposed approach improves the average link fidelity to a maximum of 0.88, compared to 0.78–0.86 achieved by existing methods. The end-to-end latency is reduced from 104 ms to 83 ms, while baseline schemes remain above 110 ms under the same conditions. In addition, the average response time is reduced to approximately 147 ms, resource utilization increases to 71–83%, and the success ratio improves from 93% to 97%, resulting in improved end-to-end reliability. The cumulative reward stabilizes at around 16.0, with convergence observed within 300 training episodes, and the achieved throughput reaches 118–121 entangled pairs per second, confirming the effectiveness of the proposed framework in dynamic NISQ network scenarios.
Distributed quantum computing provides a scalable route for executing quantum circuits beyond the capacity limits of a single quantum processing unit (QPU), but it introduces a communication-aware compilation problem involving strict hardware constraints and circuit dependencies. This paper presents an architecture-aware reinforcement-learning framework that formulates distributed quantum compilation as a constrained Markov Decision Process (MDP). The compiler-level communication actions dynamically update logical-qubit placement and enable subsequent gate execution. A heterogeneous graph model represents interactions among hardware, logical qubits, and circuit operations, while a policy trained via Proximal Policy Optimization optimizes EPR-pair consumption and communication makespan. Evaluation across benchmark circuits shows that our policy matches state-of-the-art heuristics on structured workloads, with lookahead reward shaping yielding modest improvements on unstructured circuits. These results demonstrate that reinforcement learning is a flexible alternative to manual heuristics, though scalability remains a key bottleneck for practical use.
The purpose of this research is to show the capabilities that quantum computing has to offer for AI. Therefore, the scheduling problem was chosen in order to demonstrate and compare the classical reinforcement learning (CRL) agent with the quantum computing reinforcement learning (QC-RL) agent. This constrained combinatorial optimization problem involves scheduling an oil field with several drilling rigs which are limited in capacity, area limitations, mandatory maintenance periods, weather restrictions and a mobilization costs which is considered along with well priorities and deadlines. Deep reinforcement learning (DRL) can learn adaptive scheduling policies, but value-based agents such as Deep Q-Networks (DQN) are known to exhibit high variance across random initializations. This paper investigates whether a hybrid quantum-classical value function can improve this reliability. We formulate realistic multi-rig drilling scheduling as a Markov Decision Process with action masking over an eight-constraint simulator (24 wells, 6 rigs, 3 areas, 80-day horizon) and compare a classical DQN against a hybrid Quantum DQN in which the value network’s core is a variational quantum circuit via a 4-qubit ZFeatureMap encoding followed by a RealAmplitudes ansatz with 12 trainable quantum parameters, executed using a PyTorch-Qiskit interface. Both agents share an identical training and evaluation protocol (3 seeds × 30 episodes, greedy evaluation) to ensure a controlled comparison. Under greedy evaluation, the hybrid Quantum DQN completes 22.3/24 wells on average versus 17/24 for the classical DQN, achieves an 12.87% higher mean reward (9.12M vs 8.08M), and most significantly reduces the across-seed reward standard deviation by roughly 6.8× (0.202M vs 1.38M). These results indicate that the structurally constrained quantum value function acts as an implicit regularizer, yielding scheduling policies that are markedly more consistent across initializations than a substantially larger classical network. This simulated quantum computing experiment demonstrates the potential for quantum computing in the field of AI.
This paper addresses the critical challenge of resource allocation in non-homogeneous quantum computing environments, where quantum bits (qubits) exhibit significantly varying characteristics. We propose a novel reinforcement learning-based algorithm for optimizing resource allocation, specifically designed to identify the optimal execution path for quantum computing tasks and maximize qubit utilization. The core of our approach lies in building a probabilistic model representing the quantum computing environment and employing a reinforcement learning agent to learn the best qubit allocation strategy through trial and error. This algorithm overcomes the limitations of existing methods that assume uniform qubit properties, leading to improved efficiency and success rates in quantum computations. The algorithm's core equation is as follows: * *E* = ∑i=1K *ri* (1 - *pi*) where *E* represents the overall execution efficiency, *K* is the number of qubits, *ri* is the allocated resource for qubit *i*, and *pi* is the probability of failure for qubit *i*. The algorithm's objective is to minimize *E*. Furthermore, the reward function is defined as: * R = (1 - E)* where *R* is the reward and *E* is the execution efficiency. The algorithm learns by maximizing this reward. The probabilistic model incorporates qubit characteristics such as coherence time, gate fidelity, and coupling strength, represented as probability distributions. The reinforcement learning agent utilizes a Q-learning algorithm to estimate the optimal Q-values, representing the expected reward for each state-action pair. The state space includes qubit characteristics and task requirements, while the action space consists of qubit allocation decisions. The algorithm iteratively updates its Q-values based on observed rewards, converging towards the optimal resource allocation strategy. We demonstrate that this approach offers a significant improvement over traditional uniform qubit allocation methods, particularly in scenarios with highly heterogeneous qubit populations. The primary contribution of this work is the integration of probabilistic modeling and reinforcement learning to effectively address the complexities of non-homogeneous quantum computing.