Research Projects

In impedance-type haptic interfaces, encoders are typically employed to provide high resolution position measurements from which velocity is estimated, most commonly via the finite difference method (FDM). This velocity estimation technique performs reliably, unless very fast sampling is required, in which case noise or delay due to filtering of the position signals reduces accuracy in the estimate. Despite this limitation, FDM is attractive because it is a passive process, and therefore the passivity of the overall system can be guaranteed. Levant's differentiator is a viable alternative to FDM, and exhibits increased accuracy in velocity estimation at high sample rates compared to FDM. However, the passivity of this nonlinear velocity estimation technique cannot be shown using conventional methods. In this paper, we employ a time domain passivity framework to analyze and enforce passive behavior of Levant's differentiator for haptic displays in discrete time.

Our goals in this research project are to determine the significance of performance of inanimate tasks as a marker for robotic proficiency and assess the utility of inanimate task training on robotic skill performance.  We aim to establish standardized tasks for training, define accurate metrics for performance, and assess motor skill acquisition in virtual and real environments.

Robot-assisted surgery offers distinct advantages and is rapidly being applied to a diverse range of surgical procedures.  However, teleoperation inherently decouples the surgeon from the patient.  While robotic-assistance permits a more natural, intuitive interface in comparison to standard laparoscopy, there is still a significant learning curve in mastering the technique.  In addition, the advantages of the robotic system are further limited by the lack of tactile and kinesthetic information transmitted to the surgeon.  Given this lack of sensory feedback, more emphasis is placed on interpreting visual cues and understanding robotic movement during performance.

As yet underdeveloped is the psychology of human learning as it pertains to manual control tasks in fully dynamic, multi-degree-of-freedom domains. While we currently possess the capacity to teach these tasks, we are unable to predict how well people will do in these domains or how rapidly they will learn.

This project studies human performance and acquisition of sensorimotor tasks in real and virtual environments. Human motion data and performance of various skills by performers who exhibit linear performance gains are being analyzed and compared to data for subjects who rapidly acquire skill and exhibit nonlinear performance gains. This data will lead to development of more accurate models of sensorimotor skill acquisition, and doing this should lead to improved understanding of training methods in human motor learning domains.

Sensing of displacement using only inertial measurement devices (IMDs) such as rate gyros and accelerometers is an active research topic with many diverse applications in biomechanics, human motion, earthquake engineering, robotics and mixed reality interfaces.

The main challenge in using integration to obtain velocity or position data from IMDs is DC drift in measurements. Even very small DC offsets in acceleration measurements lead to significant errors that increase linearly in single integration and parabolically in double integration. This severely complicates the problem of sensing static (DC) displacements using IMDs. Nevertheless, a significant amount of literature has focused on improving absolute (both static and dynamic) displacement estimations in spite of DC offsets.

Interest in the rehabilitation applications for robots has been increasing (Erlandson 1992, Reinkensmeyer et al. 1996, Reinkensmeyer et al. 2000).Considering their many advantages -such as allowing repeability and scalability and safety- it can be said that  rehabilitation robots are the most efficient and effective means to help SCI patients.