Transparency is a key performance evaluation criterion for haptic devices, which describes how realistically the haptic force/torque feedback is mimicked from a virtual environment or in case of master-slave haptic device. Transparency in haptic devices is affected by disturbance forces like friction between moving parts. An accurate estimate of friction forces for observer based compensation requires estimation techniques, which are computationally efficient and gives reduced error between measured and estimated friction. In this work different estimation techniques based on Kalman filter, such as Extended Kalman filter (EKF), Iterated Extended Kalman filter (IEKF), Hybrid extended Kalman filter (HEKF) and Unscented Kalman filter (UKF) are investigated with the purpose to find which estimation technique that gives the most efficient and realistic compensation using online estimation. The friction observer is based on a newly developed friction smooth generalized Maxwell slip model (S-GMS). Each studied estimation technique is demonstrated by numerical and experimental simulation of sinusoidal position tracking experiments. The performances of the system are quantified with the normalized root mean-square error (NRMSE) and the computation time. The results from comparative analyses suggest that friction estimation and compensation based on Iterated Extended Kalman filter both gives a reduced tracking error and computational advantages compared to EKF, HEKF, UKF, as well as with no friction compensation.
Unscented Kalman Filter for State and Parameter Estimation in Vehicle Dynamics
