An IMU-driven rider-on-saddle detection system for Electric Power Assisted bicycles

The paper addresses the design of a rider on saddle detection algorithm for an all-in-the-wheel Electric Power Assisted bicycle that is not equipped with a torque sensor. The algorithm is based on inertial and motor velocity measurement along with pedal cadence. The proposed system is based on the hierarchical analysis of three different dynamics: longitudinal, lateral and vertical dynamics. By framing the problem as a fault detection system, it is shown that the system can activate the assistance only when the cyclist is on the saddle, thus increasing safety.