Abstract
In this paper, a sensor fusion algorithm is proposed for electric bicycles to accomplish power-assist function without using torque sensors. The sensor fusion is observer-based and uses outputs from the wheel encoder and a 6-axis inertial measurement unit to estimate the longitudinal acceleration of the bicycle and the slope angle of the road. It is mainly based on the kinematic model that describes the time-varying characteristics of the gravity vector in a moving frame. By exploiting the structure of the observer model, convergence of the estimation errors can be easily achieved by selecting two sub-gain matrices in spite of the time-varying characteristics of the model. The fusion results allow one to conduct mass compensation, gravity compensation and friction compensation for power-assist purposes. With the compensations, riding the power-assist bike on hills is similar to riding a conventional bicycle on the level ground regardless the weight increase by the battery and the motor.