Accurate and robust state estimation for bicycles

Abstract

Accurate estimates of the dynamical states of bicycles are crucial for many advanced rider assistance systems. However, systems that can provide an exact estimate of the system states (especially the system orientations) are often expensive and therefore cannot be used in mass production. In this work, a method is presented that estimates the dynamical states of a bicycle using measurements, provided by a cost-effective sensor configuration. The proposed method is based on a constrained extended Kalman filter and uses accelerometer, gyroscope and wheel speed measurements to estimate the vehicle dynamics. Since no bicycle model is required, the filter can be easily adapted for the use in a wide range of bicycles and other single-track vehicles like motorcycles. The filter is implemented on a rapid control prototyping platform and the results are compared to measurements of a reference sensor unit. Here, special attention is put on the roll angle and the velocity estimates, where the filter produces excellent results. In addition, the filter robustness to sensor errors and uncertain system parameters is evaluated.