Medical Assistance Systems

Our joint paper "Knowledge- and memory-based interactive task learning for human robot collaboration" appeared on 16 September 2026 in Frontiers in Computer Science (Section: Human-Media Interaction). It was written as a collaborative work of the AICOR Institute for Artificial Intelligence, the Medical Assistance Systems research group and the Italian Institute of Technology (IIT). 

The paper addresses task generalization. Robots trained by imitation reproduce observed motions but not the semantics behind them, so a robot that has learned to slice a cucumber fails when asked to quarter an apple. The framework links actionable knowledge graphs (AKGs), which hold object roles, affordances and parameter ranges such as cut position or tilt angle, with narrative-enabled episodic memories (NEEMs), which record concrete interactions. Shared identifiers connect the two, so the robot can query the parameter space of an observed action and ask for missing values.