WP2
AI for Monitoring, Detection & Adaptive Decision-Making (AUTH)
Challenge
Robots often rely on deep learning models that can be overconfident and reactive, making them unreliable in unexpected situations where safety matters most.
Our Tools
- Uncertainty quantification and anomaly detection.
- AI-based decision support linked to digital twins.
- Neural architectures from CNNs to LLMs for knowledge update.
- Open-source software following EU Open Science principles.
How it fits
WP2 provides the “intelligence layer” of RoboSAPIENS – detecting when the system is unsure and ensuring safe replanning within the MAPLE-K loop.
Progress so far
- 8 open-source tools developed.
- Evaluation ongoing across all use cases.
- Foundation model and LLM innovations influencing the project strategy.
Impact & What’s Next
These tools make robots more self-aware and trustworthy.
Next: deepen integration with use cases and deliver generic AI modules ready for industry.
Technology, Tools & Architecture
RoboSAPIENS delivers the core technologies needed to make robots truly self-adaptive – reacting to change without sacrificing safety, trustworthiness, or performance.
Our research partners provide a complete ecosystem of tools that work together across the robot’s lifecycle:
WP1 ensures adaptation is correct by design through modelling and verification.
WP2 gives robots awareness – detecting uncertainty and anomalies.
WP3 keeps operations trustworthy in real time.
WP5 turns innovation into deployable robotic systems.
Together, these technologies enable robots that don’t just follow instructions – they think, assess, adapt, and stay safe.
From industrial assembly to maritime navigation, RoboSAPIENS is making autonomy more robust, scalable and human-centred.