Verification of sinusoidal steady state system identification of a Phantom Omni haptic device using data driven modeling

Abstract

Haptic feedback has two important sources of dynamics: the machine being controlled and the haptic device itself. This paper concentrates on the means of identifying the dynamics of a Phantom Omni haptic feedback device. Two models are compared: a dynamic model with parameters using results from sinusoidal steady state analysis and a data driven model that uses pseudo-random binary sequences (PRBS) for identification. The overall form of the frequency and phase response is well-defined for the dynamics model but for the data driven model a spectral estimate from PRBS response data is used to determine the model order. The results in this paper show that a dynamic equation based minimal model produces accuracy as good as the data driven model. While the data driven model has more fitting accuracy the increase in accuracy is not useful for modelling the physical response as the differences occur at high frequencies where the Phantom arm is not sensitive anyway. The dynamic model is particularly useful as it gives a physical basis for the observed output and the sinusoidal steady state behaviour is useful for exposing non-linearities. Future work includes development and verification an arm inertia model that allows system parameters to be identified from response data at arbitrary arm angles.