Motorcycle longitudinal and lateral state estimation via Kalman filtering

Abstract

Motorcycle safety systems rely on accurate state estimation of vehicle quantities. Systems like Traction Control (TC), Anti-lock braking system (ABS) and anti-wheelie (AW) are based on knowledge of vehicle states related to both longitudinal and lateral dynamics. In particular, cornering ABS and cornering TC relies on combined longitudinal and lateral dynamics. In this paper an accurate state and parameters estimator is presented, that can be used with standard sensor sets in commercial motorcycles. The estimator is based on a complex motorcycle dynamical model, with measurements coming from Inertial Measurement Unit (IMU) and wheel encoders. The estimator is based on an Unscented Kalman Filter and is tested in a realistic simulative scenario, under noisy sensors, model mismatches, and unknown initial conditions. The estimator is compared at the end with a simplified version.