A good example of the design process and evaluation of a haptic interface can be found in [12], where Ellis et. al. discuss the design and performance of a 3 DOF planar haptic interface. They discuss human operator constraints (such as workspace, position and force bandwidth), mechanical constraints (inertia, friction and stiffness), and environmental constraints (actuator choices and power requirements). They evaluated the structural dynamics of their device, along with the force bandwidth and PI controller stiffness. Together, these criteria make a powerful tool for evaluating the performance of a haptic display.
In [19], Jex describes four simple virtual environments that together form a test for a haptic display. If the display is able to successfully implement all four environments, then there is a high probability that it can successfully implement any virtual environment. Jex's four benchmark virtual environments are :
This work is useful in that it establishes a set of goals that must be achieved in order to indicate that a device can perform suitably. In this thesis, we incorporate the first three tests into a virtual wall, where the operator should feel unconstrained outside the wall, unilaterally constrained inside the wall, and have the device come to rest in a natural way if the operator lets go.
In [26] Rosenberg and Adelstein examined how operators perceived virtual walls, decomposing wall perception into three qualities : crispness of initial contact, hardness of surface rigidity, and cleanness of final release. Their experimental results indicated that high stiffness is necessary for operators to perceive hardness, while damping is necessary to perceive crispness, with stiffness playing a more important role in overall "wall-like perceptual character." They also found that their subjects preferred directional damping (damping that is active only in one direction) to conventional, as it led to a cleaner release from the virtual wall. This type of research is likely to be a necessary part of haptic display in the years to come, since subject impressions of virtual environments are critical. For any given physical system, there are several different ways to model its physical behavior. By perceptually decomposing the environment, and comparing the implementations based on different models, we can determine which model users perceive to be the most effective and why.
In [21], Lawrence and Chapel used a haptic display to measure the limitations of human operator motion and perception capabilities. Using the results of these experiments, they put bounds around the impedances that a device needs to be able to emulate in order to simulate both hard surfaces and unrestrained motion. For example, their results indicate that a human operator cannot haptically perceive the difference between a wall of stiffness 104 N/m and a wall of infinite stiffness. The relation to our work is that while our group focuses on how to build devices with a greater dynamic range, their work tells us how wide we need to make it. While these results only apply to a device with an interface that matches the one used for the experiments, one can imagine that interaction with virtual environments would use a finite number of interfaces, so that these numbers could be obtained for each configuration.
We feel that this thesis also makes an important contribution to the field of haptic display, namely that haptic interfaces can be evaluated in terms of "Z-width," the range of impedances which a device can emulate passively. Like the performance criteria presented by Ellis et. al., we can use the concept of Z-width to evaluate and compare different device configurations, both theoretically and experimentally. Coupled with performance criteria that measure operator perceptions of virtual environments, we feel this tool could be extremely useful in the design and implementation of virtual environments.