Characterizing and controlling the high-frequency dynamics of haptic interfaces.

摘要

Haptic interfaces attempt to recreate the feel of real objects for telerobotic or virtual interactions, allowing the user to touch distant or unreachable environments and computer-generated models through a lightweight robotic arm. Current haptic rendering techniques, which use position or force feedback, generally cannot convey a crisp contact with a hard object, nor can they convey the fine features of a textured surface; instead, portrayed objects feel overly soft and unnaturally smooth or oscillatory. This work has developed methods for stably endowing impedance-type haptic interfaces with the high-frequency (20 to 1000 hertz) feedback signals necessary to make virtual and remote objects feel nearly indistinguishable from their real counterparts.; The fundamental insight for this undertaking is that haptic systems have internal electrical, mechanical, and biomechanical dynamics that strongly influence their performance. These dynamics can be modeled by careful application of developed identification techniques, including comprehensive evaluation and successive isolation, and the resulting models can be used to improve interaction realism in two main ways. First, the controller can precisely create high-frequency fingertip accelerations during contact with remote or virtual objects by inverting the interface's dynamics before playback. For teleoperation the target accelerations are measured in real time at the remote manipulator, and in virtual environments they are pre-recorded. High-frequency acceleration matching creates vibrations at the user's hand that closely correspond to the specified signals and that feel almost identical to the real object, as confirmed by a human subject study.; Second, the dynamic relationship between haptic feedback command and measured device position can be estimated and canceled to improve the stability of the interface. A position-force teleoperation system that vibrates unnaturally during contact with hard objects behaves well if induced master motion is canceled from the remote robot's movement command, eliminating distracting signals and allowing the user to feel the remote environment more clearly. Application of both of these techniques to minimally invasive surgery and medical simulation is specifically promising, as it would allow physicians to feel the hardness and texture of the structures being manipulated, potentially facilitating new procedures and improving patient outcomes.