The Trustworthy Autonomous Systems (TAS) group at IDSIA focuses on developing AI systems that are safe, reliable, explainable, and aligned with human values. Our research bridges theoretical foundations with real-world applications in autonomous mobility, critical infrastructure, and human-AI interaction.
Research Areas
Click on each area to expand and learn more.
Explainable AI (XAI)
We develop methods to make AI decision-making transparent and interpretable. Our work includes surrogate causal models, opacity taxonomies, and human-centered explanations for machine learning systems.
- Surrogate causal models for ML explanation
- Taxonomies for AI opacity
- Human-centered explainability
- Social attribution in AI systems
Related projects: MaLESCaMo, B4EAI | Team: A. Termine, A. Facchini, A. Antonucci
Trustworthy Machine Learning
Research on making machine learning systems reliable, robust, and formally verifiable. We combine probabilistic methods with formal verification techniques to ensure AI systems behave as expected.
- Imprecise probabilistic models
- Formal verification of neural networks
- Robust model checking
- Uncertainty quantification in AI
Related projects: Hasler Foundation Verification Project | Team: A. Antonucci, N. Sharygina, A. Termine
Autonomous Systems Safety
Ensuring safety in autonomous vehicles, wheelchairs, drones, and railway systems. We develop perception systems, safety assessment methods, and risk evaluation frameworks for cyber-physical systems.
- Safe perception for autonomous wheelchairs
- Drone-assisted mobility for people with reduced mobility
- Railway safety and automation
- Cyber-physical systems security
- Multi-sensor fusion for robust perception
Related projects: REXASI-PRO, AutoMoTIF, Academics4Rails | Team: F. Flammini, F. Corradini, C. Grigioni, J. Guzzi
AI Ethics
Methodologies for assessing AI trustworthiness and addressing ethical challenges in artificial intelligence. We are affiliated with the Z-Inspection® initiative for trustworthy AI assessment.
- Trustworthy AI assessment methodologies
- Ethical AI frameworks and governance
- Epistemic challenges of AI systems
- Human-AI interaction and complementarity
Related projects: B4EAI, Z-Inspection affiliation | Team: A. Facchini, A. Ferrario, A. Termine
Smart Transportation and Railways
AI applications for intelligent transportation systems, focusing on safety, anomaly detection, predictive maintenance, and digital twins for railway infrastructure.
- Smart railway systems and ERTMS
- Anomaly detection in transportation networks
- Digital twins for infrastructure monitoring
- Predictive maintenance using AI
- Conflict detection in IoT-enabled systems
Related projects: AutoMoTIF, Academics4Rails, PhDs EU-Rail | Team: F. Flammini, M.J. Pappaterra
For more details on specific research topics or collaboration opportunities, please contact the relevant team members or visit our Publications and Projects pages.