This dissertation presents a real-time haptic rendering strategy with high fidelity, while guaranteeing the stability of the haptic system. Haptic systems employ physics-based deformation models such as finite-element models and mass-spring models. These physics-based deformation models for high fidelity have to deal with complex geometries, material properties, and realistic behavior of virtual objects. This incurs heavy computational burden and time delays so that the reflective force often cannot be computed at 1 kHz which is a safe frequency for stability of the haptic systems. Lower update rates of the haptic loop and the computational delays also deteriorate the realism of the haptic system.
The dissertation proposes an output-estimation method with reduced multirate sampling for real-time MIMO (multi-input multi-output) haptic rendering with high fidelity. The haptic system includes both the graphics loop with a low update rate (JT) and the haptic loop with a high update rate (T), where J and T are some positive integer, and simulates haptic and graphic interaction involving deformable objects. Dynamics of the physics-based deformation is captured in a discrete and deterministic input-output model. The MIMO output-estimation method is developed drawing on a least-squares algorithm and an output-error estimation model. The P-matrix resetting algorithm is also designed to deal with the changing input-output relationship of the deformation model. The parameters of the discrete input-output model are adjusted on-line. The estimated parameters are not supposed to converge to static values, but rather continuously trace the real parameters of the input-output relationship of the deformation model. Inter-sample outputs are computed from the estimated input-output model at a high rate, and traces the correct output computed from the deformation model. The computational time-delay, JT, is compensated by the JT-step-ahead output model. This method enables graphi...