Identification and Rendering of Contact Dynamic Models in Robotic and Haptic Systems

Abstract

Haptic simulation systems allow kinesthetic interaction between users and virtual environments. These environments are represented using mathematical models which simulate the contact dynamics of real objects. However, the discrete and quantized nature of conventional haptic systems limits the mechanical impedance of the virtual environment and thus degrades the realism of the simulation. Digital haptic systems are particularly limited in their ability to render highly damped environments. Analog control is proposed as an alternative for implementing high impedance virtual environments, and a study of analog haptic stability is presented. A mixed analog-digital is also proposed which uses nonlinear analog feedback and real-time parameter update algorithms to render complex and multilayer analog environments. The accuracy of environment models is critical for realistic haptic simulations. Soft environments with limited deformation have been shown to be more accurately described using the nonlinear Hunt-Crossley model than linear models. The thesis also proposed a novel real-time Hunt-Crossley identification algorithm based on a polynomial approximation. Experimental results point at superior accuracy and convergence rate as compared to existing Hunt-Crossley estimation algorithms. The proposed identification method can also be used to facilitate robotic contact tasks.