Apparatus and method for fault diagnosis and back-up of advanced driver assistance system sensors based on deep learning

A method for fault diagnosis and back-up of a plurality of advanced driver assistance system (ADAS) sensors based on deep learning, in a multi-modal sensor network that combines multi-type sensors, the method comprising:

• compressing an output of one sensor of the plurality of sensors into feature vectors;
• measuring an individual reliability of the one sensor based on each of the feature vectors;
• measuring a dynamic reliability of the one sensor on the basis of a dynamic relationship between the feature vectors generated consecutively;
• quantitatively evaluating a value of a reliability of an output of the one sensor, on the basis of the individual reliability and the dynamic reliability, to be within a predetermined range of values;
• extracting a shared representation between other sensors of the plurality of sensors and quantitatively evaluating a value of a normal-operation reliability of the output of the one sensor under an operation without faults on the basis of the extracted shared representation;
• quantitatively evaluating a final reliability of the one sensor on the basis of the value of the reliability of the output of the one sensor and the value of the normal-operation reliability of the output of the one sensor;
• and
• extracting, based on the value of the reliability of the output of the one sensor, a shared representation between the one sensor and another sensor of the plurality of sensors by reducing a weight of the one sensor so that an influence of the one sensor is minimized when determining reliabilities of sensors of the plurality of sensors other than the one sensor.