AI4SDV

Engineering teams face growing challenges in:

• Managing large volumes of development and operational vehicle data.
• Ensuring software quality, safety and compliance throughout the lifecycle.
• Automating testing and validation processes.
• Reducing development effort while accelerating innovation.
• Maintaining trust and traceability in AI-assisted engineering workflows.

Traditional engineering approaches alone are no longer sufficient to address this complexity. New AI-assisted methods are needed to help engineers analyse information faster, automate repetitive tasks and make better-informed decisions.

AI4SDV is a European research and innovation project that explores how artificial intelligence can enhance the development of Software-Defined Vehicles (SDVs).

The solution will enable engineering teams to transform large volumes of heterogeneous vehicle and development data into reliable, interpretable and actionable insights throughout the SDV lifecycle.

Key innovations include:

1. Intelligent data curation pipelines. Automated pipelines will improve data quality through the following capabilities that will help ensure that AI systems are trained and operated using trustworthy and consistent data:
   1. Data ingestion and integration.
   2. Cleaning and normalization.
   3. Duplicate detection.
   4. Provenance tracking.
   5. Bias and data quality assessment.
   6. Reproducibility monitoring.
2. Hybrid AI-powered data intelligence. A hybrid approach to support more robust engineering decisions while reducing the risks associated with purely black-box AI solutions. The platform will combine:
   1. Rule-based reasoning.
   2. Statistical and exploratory data analysis.
   3. Anomaly detection techniques.
   4. Machine learning models.
   5. Generative AI capabilities.
3. Explainable decision support. By leveraging explainable AI, Retrieval-Augmented Generation (RAG) and agentic AI approaches, the system will adapt recommendations to different user roles and engineering contexts. AI recommendations will be accompanied by interpretable diagnostics that help engineers understand:
   1. Data quality issues.
   2. Potential bias patterns.
   3. Detected anomalies.
   4. Compliance-related risks.
   5. Validation bottlenecks.
4. Continuous improvement through human feedback. Engineers will remain in control through human-in-the-loop mechanisms that allow feedback to continuously improve decision-support capabilities while maintaining transparency and accountability.
5. AI-assisted Testing, Verification and Validation. A second contribution focuses on enhancing Verification and Validation (V&V) processes through AI-powered automation. Testing activities in Software-Defined Vehicles generate significant effort throughout development cycles. AI4SDV investigates how AI can automate and accelerate these activities while maintaining engineering rigour and traceability. The project will support:
   1. Automated test generation from requirements and engineering models.
   2. Test prioritisation and optimisation.
   3. AI-assisted analysis of test results.
   4. Anomaly and defect detection.
   5. Adaptive testing strategies.
   6. Validation workflow automation.

These capabilities aim to reduce manual effort, increase test coverage and improve software reliability.

Expected benefits include:

• Faster software development cycles.
• Reduced engineering effort.
• Improved software quality.
• More efficient testing and validation.
• Greater transparency in AI-assisted engineering decisions.
• Stronger trust in AI technologies for safety-critical environments.

The project ultimately contributes to strengthening Europe’s leadership in Software-Defined Vehicle innovation and next-generation automotive engineering.