A method for training a ML model to assess two- or three-wheeled vehicle driver behavior using multi-sensor fusion and relative motion analysis.

Patentradars sammanfattning

Autoliv Development AB discloses a computer-implemented training pipeline for a machine-learning model targeting driver behavior assessment in two- and three-wheeled vehicles. The method automatically collects multi-sensor data from at least two sensors placed at different positions—including at least one sensor on the driver—across multiple time-windowed episodes. Training associates each sensor-data set with a predetermined driver behavior by comparing sensor streams and computing relative motions, ultimately linking identified postures and behaviors to a degree of safety.

AI-genererad och redaktionellt processad.