Pedaling torque estimation in pedal-assist electric bicycles without external-load identification

Abstract

A new strategy for estimating the pedaling torque produced by a cyclist on a pedal-assist electric bicycle based on an unknown-input observer is proposed in this work. Estimation is carried out from a nonlinear model of the vehicle longitudinal dynamics using a disturbance observer. The main contribution is that estimation does not require either measuring or estimating the influence of constant or slowly varying loads on the vehicle, such as gravitational force or resistive torque. As another contribution, we demonstrate that the observer itself provides an estimate of the mean pedaling torque, which serves as a suitable, slowly varying signal for generating the reference torque for the electric motor in a pedal-assistance scenario. Experimental results under real operating conditions are presented using an electric bicycle equipped with a torque sensor integrated into the bottom-bracket axle.