A Human-in-the-Loop Confidence-Aware Failure Recovery Framework for Modular Robot Policies

Researchers developed a human-in-the-loop framework that helps modular robots recover from failures more efficiently. The system uses calibrated uncertainty estimates and human intervention cost models to decide which module failed and when to ask for human help. In tests with a robot-assisted bite acquisition system involving people with mobility limitations, the framework improved recovery success while significantly reducing user workload compared to traditional methods.