Quantum Software Engineering: Roadmap and Challenges Ahead

1 INTRODUCTION

• Quantum computers have demonstrated utility in approximation methods such as Variational Quantum Algorithms, while research focuses on more stable QPUs beyond the NISQ era [25, 136].These developments indicate that industrialized quantum software is approaching.Source: §1 · p. 2
• Industrial adoption depends on software that is repeatable, efficient, maintainable, reusable, and cost-effective, qualities quantum software currently lacks established procedures and standards to support.The paper argues that sound engineering principles are needed to bridge this gap.Source: §1 · p. 2
• Transferring classical engineering practices to quantum software is challenging because the field needs both adapted classical methods and new approaches for quantum requirements.Anticipated hybrid quantum-classical software also requires technologies for seamless interaction between computing environments.Source: §1 · p. 2
• Quantum testing faces challenges arising from quantum states, while requirements engineering remains less emphasized because real-world applications are limited.The paper says requirements engineering will diverge from classical approaches.Source: §1 · pp. 2–3
• The paper surveys recent QSE developments and identifies primary challenges across its research areas to guide future work.It aims to support robust, efficient, and maintainable quantum software.Source: §1 · p. 3

2 FUNDAMENTALS OF QUANTUM COMPUTING

• Qubits are quantum information units, and their superposition and entanglement provide computational properties distinct from classical systems.The section introduces these concepts as foundations for understanding later material.Source: §2 · p. 3, §2.1 · p. 3

2.1 Qubit

• For a qubit, |ψ⟩ denotes its state, while α and β are complex probability amplitudes for finding it in basis states |0⟩ and |1⟩, respectively; their squared magnitudes sum to 1.The basis states are analogous to classical 0 and 1.Source: §2.1 · pp. 3–4

2.2 Superposition

• Superposition allows a qubit to represent multiple states, but measurement collapses it to a classical state determined by the probability amplitudes.For n qubits, the number of representable states grows as 2^n.Source: §2.2 · p. 4

2.3 Entanglement

• Entanglement links qubit states so that measuring one qubit reveals information about its partner, regardless of spatial separation.The paper says entanglement enables certain quantum algorithms to achieve exponential speed increases that are impossible without it.Source: §2.3 · p. 4

2.4 Quantum state vector and quantum register

• A quantum state vector represents a system’s configuration during algorithm execution; for two qubits, its amplitudes correspond to four basis states.Measurement probabilities are determined by the squared moduli of those amplitudes, and larger systems can group qubits into quantum registers.Source: §2.4 · pp. 4–5

2.5 Circuit-based quantum computing and Quantum annealing

• Circuit-based quantum computers run general-purpose algorithms, whereas quantum annealers specialize in optimization problems.The paper focuses its fundamental concepts on circuit-based computers while also discussing annealing work.Source: §2.5 · p. 5

2.6 Quantum Gate

• A quantum gate manipulates qubits and is represented by a 2^n × 2^n unitary matrix for a gate affecting n qubits.Examples include Hadamard for superposition, Pauli-X for flipping qubit states, and CNOT for controlled operations that create entanglement.Source: §2.6 · p. 5

2.7 Quantum Circuit

• A quantum circuit orders gates, measurements, and potentially classical logic to carry out a computation on quantum states.The illustrated four-qubit circuit produces numbers less than three: 0000, 0001, and 0010.Source: Figure 1 · p. 5, §2.7 · p. 6

2.8 Quantum Algorithm

• Shor’s algorithm uses period-finding and the Quantum Fourier Transform for an exponential factoring speedup, while Grover’s algorithm gives unstructured search a quadratic speedup through amplitude amplification.The section also notes the potential implications of Shor’s factoring capability for RSA encryption.Source: §2.8 · p. 6

2.9 Quantum Oracle

• An oracle is a circuit that transforms an input quantum state and supplies the resulting state to another circuit implementing an algorithm.The paper identifies oracles in algorithms including Grover, Deutsch-Jozsa, Simon, and Bernstein-Vazirani.Source: §2.9 · p. 7

2.10 Quantum Noise

• Quantum noise can arise from environmental interactions, gate and measurement errors, cross-talk, and qubit relaxation or dephasing.Simulators can also introduce noise artificially to test error-mitigation and correction strategies.Source: §2.10 · p. 7

QSE’s development

• The paper traces QSE’s early recognition to 2002 and notes that IBM’s 2016 release of a gate-based quantum computer and online simulator preceded growth in the field.Clark et al. framed QSE as a grand challenge in computer science research.Source: §3 · p. 7

Research venues

• QSE research is represented in dedicated conferences, workshops, and journal venues, including Q-SE, QSW, Q-SET, and QCE.The paper also identifies dedicated journal issues or sections in IEEE Transactions on Quantum Engineering, ACM Transactions on Quantum Computing, Journal of Systems and Software, Information and Software Technology, and ACM Transactions on Software Engineering and Methodology.Source: §3 · p. 8

Publication trends

• More than 200 QSE publications were recorded in 2023, following increased publication activity after quantum simulators and platforms became available around 2020.The paper describes earlier publications from 2002–2007 and a resurgence beginning around 2013.Source: §3 · p. 8, Figure 2 · p. 9

Research publication venues

• Of 102 Scopus-indexed QSE papers through 31 July 2024, 70 (68.6%) were conference publications, 26 (25.5%) journal articles, and 6 (5.9%) book chapters.The authors identify conferences as the main venue for QSE research papers.Source: §3 · p. 8

Publication trends

• Among 26 journal publications, 10 appeared in the first half of 2024.Source: §3 · p. 9

Publication topics

• Quality Assurance leads the topic counts with 16 publications, followed by Service-Oriented Computing with 12.Programming Paradigms and Software Development Processes each have 9 works; Model-Driven Engineering has 8, Software Architectures 7, and Artificial Intelligence 3.Source: §3 · p. 9

• The section surveys selected QSE work and researcher opinions to identify challenges anticipated in the coming years, rather than conducting a systematic analysis.Source: §4 · p. 9

Testing and debugging landscape

• Quantum testing and debugging address functional correctness and failure diagnosis, but noise, limited hardware or simulator resources, and limited observability complicate both.Prior work spans coverage criteria, advanced testing techniques, assertions, mutation testing, and testing frameworks.Source: §4.1 · pp. 9–10

Efficient test oracles

• Most test oracles are costly, may need generation for each input-output pair, and do not scale well with scarce quantum resources.State tomography is computationally expensive, and support for checking program states during execution remains cumbersome; the section calls for more efficient oracles and empirical guidance on balancing runs and reliability.Source: §4.1 · p. 12

Test scalability

• Current test-data generation methods struggle to cover the high-dimensional input space of complex quantum software, including entangled and superposition states.The section calls for methods that detect faults in non-trivial programs while accounting for resource limits and representative test data.Source: §4.1 · p. 12

From simulators to real quantum computers

• Testing on noisy hardware can make results unreliable, while limited access means current solutions mainly rely on simulators.The authors call for noise reduction in testing and test optimization to make future real-device testing cost-effective.Source: §4.1 · pp. 12–13

(Quantum) Artificial Intelligence and (Quantum) Software Testing

• Classical AI may support quantum testing tasks such as repair, mutant generation, oracle selection, and test-case generation, while quantum AI has been applied to classical testing tasks.The paper gives examples including QAOA for test-case optimization and quantum annealing for test-case minimization; practical use requires formulating the task as a mathematical problem efficiently solvable on a quantum computer.Source: §4.1 · p. 13

Other aspects of testing

• Open issues include whether artificial faults represent real faults, mutation-operator subsumption, testing across phases, and replicable experiments.The section also calls for repository mining to identify pain points and best practices, and notes proposed testing principles and reporting guidelines.Source: §4.1 · pp. 13–14

Quantum service-oriented computing

• Classical service-oriented approaches do not transfer directly to quantum software, and prior work has explored deployment, orchestration, workflows, API patterns, DevOps, and distributed quantum execution.Examples include TOSCA-based deployment, hybrid-workflow orchestration, workflow execution of variational algorithms [17], and distributing shots across QPUs.Source: §4.2 · pp. 14–15

Interoperability

• Standardized frameworks and QPU APIs are needed to connect diverse hardware, software stacks, and development environments; intermediate representations such as QIR can support portability.OpenQASM illustrates the value of standardization, while QIR is proposed as a bridge between high-level languages and hardware-specific instructions.Source: §4.2 · p. 15

Platform independence

• Vendor-specific hardware, cloud environments, and optimizations can force developers to tailor applications to platforms, increasing vendor lock-in and limiting flexibility.Intermediate abstraction layers, universal compilers, and vendor-neutral SDKs are proposed as building blocks for platform independence.Source: §4.2 · pp. 15–16

Demand and Capacity Management

• Hybrid workflows require coordination of classical and quantum resources, while capacity planning must account for hardware limits, quantum-service availability, queuing, latency, reliability, and cost.Pricing may depend on processor time, qubit count, and circuit complexity, motivating capacity and pricing strategies suited to quantum services.Source: §4.2 · pp. 16–17

Workforce training

• Moving developers from classical service-oriented architecture to quantum service-oriented computing requires both new technical skills and methods for integrating quantum principles with classical architectures.Proposed responses include progressive training and hands-on learning, alongside tools and methodologies that provide familiar interfaces for building, testing, and deploying hybrid quantum services.Source: §4.2 · pp. 17–18

Prior work

• Existing work spans quantum code generation and verification, model-driven optimization and circuit design, modernization and reverse engineering, and modelling standards and domain-specific languages [129].The section also describes UML, BPMN, and architecture-description-language approaches to modelling hybrid systems [139].Source: §4.3 · pp. 18–19

Modelling quantum-specific constructs

• Quantum MDE needs constructs and tools that represent gates, circuits, states, and measurement probabilities to model quantum software accurately.Existing uncertainty-modeling approximations could be explored for quantum software.Source: §4.3 · pp. 19–20

Development of high-level design methodologies

• Hybrid software needs high-level modelling frameworks that make quantum-classical interactions comprehensible and support design without requiring low-level quantum details.The section proposes developing domain-specific modelling languages for quantum-classical integration.Source: §4.3 · p. 20

Scalable quantum software maintenance and evolution

• Scalable quantum software maintenance calls for methods to predict change impact, assess compatibility and performance across versions, and support regression testing and automated refactoring.The section suggests sophisticated quantum software metric models as part of this work.Source: §4.3 · p. 20

Intelligent code generation and orchestration

• Intelligent code generation should translate high-level models into executable quantum code optimized for hardware constraints, while orchestration should coordinate hybrid components and resources.The proposed optimization considerations include qubit connectivity and gate fidelity.Source: §4.3 · p. 20

Programming languages and abstraction

• Quantum languages include imperative, functional, and declarative examples, while other work explores raising abstraction through quantum types, operations, and constraint-to-QUBO translation.Examples include QASM, Ket, and Q#; functional QML and Quipper; and declarative Forest [62].Source: §4.4 · pp. 20–21

Programming strategies and abstraction

• A quantum program manipulates a superposed state to increase solution amplitudes, yet common languages expose low-level gates operating directly on qubit phases and amplitudes.The section contrasts this with classical programs, whose steps transform machine state toward constructing a result.Source: §4.4 · p. 21

Complexity of circuits

• Wide, deep circuits strain NISQ hardware: qubit demands challenge available capacity, while long gate sequences increase noise and decoherence risks.Optimization directions include reducing circuit depth, compressing circuits, reusing qubits, and mapping to hardware connectivity [75] [120].Source: §4.4 · p. 22

Composable and reusable quantum software

• Quantum software needs reusable, composable circuits, but existing tools remain immature and composition must preserve properties such as coherence and phase relationships.Better circuit documentation, modular architectures, and flexible composition are identified as needed for reuse [154].Source: §4.4 · pp. 22–23

Abstractions for quantum software

• Quantum abstractions should move beyond classical-like types toward quantum-native data such as molecules or quantum fields, with operations that represent their interactions.The section argues these abstractions would better align programming with quantum computers’ potential for modelling and simulating quantum systems.Source: §4.4 · p. 23

4.5 Software Architectures

• Systems should be designed as whole information systems around their functions, rather than integrated merely because they use different computational paradigms.Source: §4.5 · p. 24
• The section frames four open challenges: architectural decisions, hybrid-system design patterns, empirical validation, and the evolution of hybrid architectures.Source: §4.5 · p. 24
• Architectural choices must balance hardware-aware performance, cross-platform interoperability, resource costs, and scalability as quantum-classical systems evolve.Source: §4.5 · pp. 24–25
• Low-level patterns should support modular, reusable, hardware-independent components that combine quantum circuits with classical routines.Source: §4.5 · p. 25
• High-level patterns should orchestrate quantum services and hybrid workflows, including task scheduling, resource allocation, and error monitoring.Source: §4.5 · pp. 25–26
• Research should empirically assess adapted classical patterns in industry and develop quantum-native patterns for future hardware and workflows.Source: §4.5 · p. 26
• Hybrid architecture evolution can accumulate technical debt, motivating refactoring, automated maintenance and testing, and sustained documentation and version control.Source: §4.5 · pp. 27–28

4.6 Software Development Processes

• Lifecycle proposals include waterfall adaptations, an iterative model using quantum data provenance, and an incremental commitment spiral model incorporating risk management.Source: §4.6 · p. 28
• The section identifies three process challenges: iterative hybrid development, risk management, and project management.Source: §4.6 · p. 28
• Iterative lifecycle models should support rapid quantum-technology change and co-development of classical and quantum components, with quantum-specific patterns, testing, and tools.Source: §4.6 · p. 29
• Engineering practices can mitigate some quantum-software risks, and the cited table summarizes risks alongside mitigation approaches.Source: §4.6 · p. 29, Table 1 · p. 29
• Risk management must account for uncertain adoption, scarce real hardware, simulator-to-production differences, and the need for adaptive strategies as quantum computing evolves.Source: §4.6 · pp. 28–30
• Project-management research could adapt agile toolchains and CI/CD for quantum software, while LLM code-generation tools raise accuracy, security, integration, and oversight concerns.Source: §4.6 · p. 30

4.7 Artificial Intelligence

• AI methods are being explored for circuit and qubit-allocation optimization, noise mitigation, simulation, debugging, verification, and testing.Source: §4.7 · pp. 30–31
• Examples include machine-learning selection of compiled circuits, prediction of execution waiting times, and AI applications to variational algorithms.Source: §4.7 · p. 31
• Other work covers neural-network circuit synthesis, quantum-code generation with LLMs, bug analysis in quantum machine learning, and quantum-program repair.Source: §4.7 · p. 31
• Quantum software engineering also supports quantum-enhanced AI, although loading classical data remains a challenge for these approaches.Source: §4.7 · p. 31
• The paper characterizes AI and quantum software engineering as a developing intersection with potential for innovation in both fields.Source: §4.7 · pp. 30–31
• AI-assisted quantum software research spans circuit optimization, hybrid AI-quantum workflows, error mitigation and correction, and scalable development support.Source: §4.7 · pp. 32–33

4.8 Summary of the different key challenges of quantum software engineering

• Testing, service-oriented computing, and model-driven engineering challenges cover test oracles, scalability, interoperability, platform independence, modeling, maintenance, and orchestration.Source: §4.8 · p. 33
• Programming paradigms, software architecture, and development processes address circuit complexity, reuse, hybrid design, empirical evidence, architecture evolution, iteration, risk, and project management.Source: §4.8 · p. 33
• Quantum artificial intelligence challenges include circuit optimization, hybrid workflows, error mitigation and correction, and scalable AI-assisted development.Source: §4.8 · p. 33

Artificial Intelligence for Software Engineering (AI4SE)

• AI and quantum computing may mutually support circuit design, optimization, error management, and more capable machine-learning algorithms.The section describes AI tools for quantum systems and possible quantum acceleration of AI tasks.Source: §5 · pp. 34–35

Software Engineering by and for Humans (SE4H)

• Human-centered design can make quantum applications more understandable and accessible while addressing ethical concerns and user needs.The section points to interfaces that simplify interaction with quantum machines for non-experts while retaining control and oversight.Source: §5 · p. 35

Automatic Programming

• AI-based automatic programming could generate and optimize quantum code from high-level descriptions, broadening access to quantum development.The section also describes possible reductions in human error and development time.Source: §5 · p. 36

Software Security

• Quantum software security must address risks to classical public-key cryptography and adapt secure coding, vulnerability detection, and monitoring to quantum environments.The section also identifies quantum-resistant cryptography and secure quantum communication protocols as relevant measures.Source: §5 · p. 36

Validation and Verification (V&V)

• Quantum V&V must account for probabilistic algorithms and support circuit correctness, quantum-bug detection, and reliable behavior across execution environments.The section also suggests quantum computing could accelerate complex verification tasks and software-model analysis.Source: §5 · p. 37

Sustainable Software Engineering

• Sustainability concerns include the energy required to cool and operate large-scale quantum processors and the ecological footprint of quantum systems.The section suggests optimizing algorithms to use fewer qubits or shorter runtimes, while quantum computing may also help address sustainability challenges such as energy distribution and climate modeling.Source: §5 · p. 37

• Quantum software development requires new techniques compared with classical software development, and these techniques are beginning to form the body of QSE.Source: §6 · p. 37
• As quantum computers attract industry attention, empirical software engineering methods and techniques need revisiting to meet industry expectations.Source: §6 · p. 38
• The authors identify noteworthy QSE challenges and call for further analysis of their synergies and dependencies, with hybrid software systems beyond the NISQ era as a supported target.The roadmap is expected to evolve as quantum technology continues to develop.Source: §6 · p. 38