Track 09 of 09
Trustworthy, Explainable and Safe Autonomous AI
The assurance layer: safety, security, explainability, robustness, fairness, governance, standards and law for systems that act physically.
Scope
The assurance layer: safety, security, explainability, robustness, fairness, governance, standards and law for systems that act physically.
Research topics
Topics include, but are not limited to:
- Safety cases and assurance arguments for learned components
- Runtime monitoring and safe fallback
- Formal methods for autonomy
- Adversarial robustness in the physical world
- Explainability for operators and regulators
- Uncertainty quantification and calibration
- Alignment and specification of autonomous objectives
- Cybersecurity of robotic and OT systems
- Privacy-preserving perception
- Auditing, logging and accountability
- Liability, certification and regulatory frameworks
- AI governance and standards (functional safety, AI management systems)
Emerging challenges
- Certifying non-deterministic components
- Explanations that are actionable for a plant operator
- Physical adversarial attacks
- Specification gaming in embodied agents
- Incident reporting norms
- The gap between AI ethics principles and machine-level assurance
Industry applications
- Safety certification of cobot cells
- AV safety cases
- Medical-device autonomy approval
- OT/ICS security
- Regulatory compliance tooling
- Insurance and risk assessment for autonomous assets
Keywords
Every submission must connect to physical action, embodiment or real-world deployment, use the IEEE conference format (A4, up to 6 pages including references) and include a Deployment Risk and Limitations statement (max. 200 words). Submission guidelines →