Automated Software Engineering

Automated software engineering encompasses the use of tools, methods and models to automate or assist in the full spectrum of software development activities, from requirements capture to deployment and maintenance. At its core, it seeks to reduce manual effort, improve consistency and accelerate delivery without sacrificing quality. Key approaches include model-driven engineering, which uses abstract representations (for example UML diagrams or domain-specific models) to generate or verify code; search-based techniques that frame tasks such as test-case generation or refactoring as optimisation problems amenable to metaheuristic solvers; and program synthesis methods—both deductive and inductive—that derive executable implementations from specifications or examples. Recent years have also seen the rise of data-driven automation: mining vast software repositories to infer coding patterns, applying machine-learning to predict defects or estimate effort, and integrating security checks into continuous delivery pipelines. Collectively, these advances underline a global trend towards self-service, adaptive and correct-by-construction pipelines that embed formal analysis, probabilistic guarantees and practical feedback loops to support developers and end users alike.

Research from Nature Portfolio

A recent study introduced a formal synthesis framework for discrete-time state-feedback digital controllers that jointly ensures defined step-response performance criteria and implementation robustness. By combining counterexample-guided inductive synthesis with finite-word-length modelling, the method generates controller code that provably meets settling-time and overshoot bounds, while accounting for fragility arising from digital hardware quantisation. Experimental results demonstrated that controllers satisfying both design and implementation constraints can be automatically derived and verified in a model-checking toolchain, streamlining the gap between control-system theory and deployable embedded software.

Research from all publishers

A novel model-based secure-development methodology integrates threat modelling, static assessment and targeted tests within a DevSecOps pipeline to automate security design and verification. Starting from high-level models that capture potential threats and countermeasures, the approach automatically instantiates containerised microservice case studies, executes static analyses and generates focused security tests. This end-to-end framework codifies security controls as part of continuous delivery, demonstrating effective identification and mitigation of vulnerabilities without manual expert intervention.

Another contribution proposes a language-agnostic program-synthesis engine guided by user-defined abstract domains. By specifying lightweight abstract semantics alongside concrete test cases, the synthesis tool prunes infeasible candidate programs and systematically enumerates implementations that satisfy both abstract specifications and observed behaviours. Empirical evaluation across string-manipulation and data-frame benchmarks showed parity with specialised deep-learning tools, illustrating how abstraction-driven search can scale to industrial-grade languages without bespoke solver encodings.

Automated Software Engineering publication trend

The graph below shows the total number of articles in automated software engineering across all publications each year (not limited to Nature Index journals).

Technical terms

Counterexample-guided inductive synthesis (CEGIS): An iterative approach that generates candidate programs, checks them against specifications, and refines future candidates using counterexamples from failed verifications.

Model-based engineering: A discipline whereby abstract models of software structure or behaviour serve as primary artefacts for analysis, code generation and verification activities.

Abstract interpretation: A static-analysis technique that approximates program semantics within a simplified domain to prune infeasible states and guide automated reasoning or synthesis.

State-feedback controller: A digital control algorithm that computes actuator commands by applying gains to the full state vector of a dynamic system.

Statistical model checking: A verification method that uses random sampling of model executions to estimate the probability of satisfying quantitative requirements under uncertainty.

DevSecOps: The practice of embedding security analysis and testing within the continuous integration and deployment pipeline to ensure that security controls evolve alongside functional code.

References

1. Secure software development and testing: A model-based methodology. Computers & Security (2024).
2. Absynthe: Abstract Interpretation-Guided Synthesis. Proceedings of the ACM on Programming Languages (2023).
3. Formal synthesis of non-fragile state-feedback digital controllers considering performance requirements for step response. Scientific Reports (2022).