Manuel Giménez Medina Enrique Alba
Ayesa Servicios Digitales Avanzados University of Málaga
(previously Emergya)
and
The fiQuare project pioneered the development of intelligent methodologies for automated software repair, significantly enhancing the dependability and operational resilience of IoT-based software infrastructures. By integrating artificial intelligence and advanced fault-detection paradigms, the initiative focused on automating defect localization, correction, and validation, thereby reducing human intervention and optimizing software quality. This note delineates the theoretical underpinnings, technical implementations, and empirical evaluations conducted throughout the project lifecycle, presenting key insights into its impact and scalability in real-world deployment scenarios.
The field of automatic software repair has witnessed substantial academic and industrial attention, with notable contributions leveraging semantic-driven correction techniques and generate-validate paradigms. The fiQare project extended this body of knowledge by integrating these methodologies into a domain-specific framework tailored for IoT-driven software maintenance, enhancing their applicability and efficacy within large-scale distributed environments.
The methodological approach employed in fiQare encompassed several structured phases. A comprehensive examination of state-of-the-art techniques in automated software repair was conducted through a systematic literature review. Existing defect localization methodologies were categorized and analysed to develop a taxonomy of fault-detection mechanisms. Algorithmic design and optimization efforts focused on developing AI-augmented heuristics for source-code fault detection and rectification. Experimental validation was performed through the deployment of the developed techniques in controlled environments, followed by iterative performance evaluation.
The technical realization of fiQare involved the construction of a robust, AI-powered software repair framework seamlessly integrated into IoT ecosystems. The computational infrastructure comprised high-performance computing nodes featuring dual Intel Xeon processors and Ubuntu-based deployment environments. The development toolchain utilized Java-centric repair algorithms, engineered within the Eclipse IDE and reinforced with continuous integration (CI/CD) pipelines. Automated defect correction models were implemented using AI-driven pattern recognition for dynamic patch generation and automatic validation.
The developed repair mechanisms underwent rigorous validation across multiple dimensions. Performance benchmarking assessed computational overhead, fault detection accuracy, and automated correction efficacy. Industrial-scale deployment within FIWARE Generic Enablers, such as Perseo CEP and IoT Agent UltraLight 2.0, demonstrated notable improvements in fault tolerance and system resilience. A comparative analysis against traditional manual debugging processes highlighted substantial efficiency gains and defect-resolution acceleration.
The fiQare project successfully established a novel paradigm in AI-driven software repair for IoT infrastructures much before the age of LLMs, significantly reducing the operational complexities associated with defect management. Future research directions include extending quality-and-safety by design methodologies to heterogeneous computing environments and refining predictive models for proactive fault prevention.
Acknowledgments
This research was conducted under the auspices of the fiQare consortium, comprising Ayesa, the NEO research group from UMA, and industrial partners TIGA, Ubiwhere, and Secmotic. The project was supported through funding under CDTI project code INNO-20171027 C2017/2-2.
Further information on https://www.celticnext.eu/project-fiqare/