Sensable Phantom
Classical VR haptics aims at giving the user the impression that they are touching a remote or virtual environment.
The user explores the environment by handling a knob or a stylus. The haptic device is the touch-equivalent of a visual display, like a screen or glasses showing images generated by a remote camera or a completely virtual ( VR ) scene. In this sense, the user handle or "manipulandum" is a haptic display. The main difference with a visual display is that our eyes passively view a whole scene at one glance, while our hands have to actively explore the tactile world point by point.
trade-offs
While higher quality is always desirable, the sense of presence does not require perfect visual or haptic rendering, and there will be cost compromises. A good haptic display will put its efforts into rendering those aspects which are relevant to the user, but it may well use loopholes provided by the human sensory system to avoid undue cost on aspects less noticeable, like visual presence may not require stereo except at close range
Keeping the limits of the human sensory system in mind will avoid overdesign and undue cost. Another trade-off is which parts of the visual and haptic system to replicate physically and which to render artificially. Flight simulators normally contain a physical copy of the control panel, since it is the easier way to display the instruments visually with sufficient accuracy and stability. Similarly, such simulators contain a physical control yoke and rudder pedals since it is the easier way to provide their tactile experience to the pilot. However, the aerodynamic control forces are emulated by haptic actuators, and the inertial forces on the pilot's body may be emulated by mounting the cabin on an active motion platform.
human bandwidth
Active voluntary motion of the human hand is relatively slow. It is limited to frequencies of about 3 Hz. A haptic device should definitely be able to follow such motion without much resistance, in order to feel free. But 3 Hz is not the maximum rendering frequency, because a haptic display does not render the user. Instead, it displays the physical environment like walls and objects and these can be very stiff. The question is how humans perceive this stiffness. Human perception is strongly biased toward change, and towards higher frequencies. The fingers can detect force signals of at least 300 [ Hz ]. Information about the hardness and even the roughness of objects is picked up at these frequencies. People will dislike spurious device vibrations and stick-slip at these rates.
drift
People will also dislike slow drifting motion of the position of virtual walls which should be fixed in the virtual world : once an object is "touched", position drift should be avoided. But absolute positioning accuracy is irrelevant to humans. People cannot reliably distinguish absolute positions of their hands or feet to within more than a few centimeters, and they will accept large offsets between the visual and the haptic world. We can look at a screen showing the avatar of the haptic manipulandum near to or far away from the workspace of our hands, or scaled up and down by a large factor, without objecting to or indeed noticing the fact. This can even be used to good advantage to expand the perceived workspace by secretly changing the haptic frame of reference on the fly during hand motions, as in Conti's very effective "deliberate workspace drift".
colocation
The exception to this rule is the case of colocation, where the device position needs to coincide exactly with a "see-through" graphics display, as in augmented reality. This method is exceedingly difficult, and best avoided. The forward direction from the human torso needs to be parallel to the viewing direction of the human head, but the two do not need to be parallel themselves. Humans can look into a screen to the side, and move the haptics device fore and aft from their body without problem. However, any rotational offset between the viewing direction and the forward direction from the torso is unacceptable, except to the dentist, who is used to work while looking in a moving and rotating dental mirror. After two years of training, dentists do not even notice rotational offsets anymore.
rendered bandwidth
An ideal haptics device can create a sensation all the way from totally free motion "like handling a piece of balsa wood in free air" ( Jex, 1988 ) to totally passive hard contact, like on a stiff physical object such as a wall or a table.
The range between these two extremes of "mechanical impedance" is sometimes used as a measure of quality for a haptic display. This range of impedances is called the "Z-width". In practical applications, a device need not span the whole range. Assistive devices do not need rock-hard virtual walls, and master-slave joysticks for handling heavy loads may not need to be very light.
Smooth motion however is always a hallmark of quality. Tactile rumble or vibration, as well as audible motor noises will spoil the sense of presence. Deliberate vibrations or audible clicks on virtual contact can be used to increase the impression of hardness, but otherwise they need to be absent. Humans perceive the rate of increase of the contact force on touching a surface as a measure of the surface stiffness. This is sometimes called "rate hardness".
grounding
Another subdivision is into devices which are "grounded", i.e. fixed to the floor or the table, as opposed to "ungrounded", i.e. carried on the hand or the wrist and unable to provide absolute reaction forces. Ungrounded devices can display gripping forces, but otherwise they belong to the realm of passive, gesture based interfaces, perhaps enhanced by pseudohaptics illusions. In this document we will mainly discuss classical, grounded haptics.
degrees of freedom
Most classical haptics devices are "point based" : they have three active de-grees of freedom X, Y and Z. The powered point can be a round knob, or the tip of a pen or stylus. Multi-DOF devices have been built, but they are problematic because it is not natural for humans to identify part of their body ( like their hand ) with a 6-DOF object. Grippers are a more natural interface, and these can be powered in torque as well as in position, but often their rotations are only tracked and not driven. Driving rotational DOF's is mechanically complex, and the last "throttle handle" or "stylus rolling" motion often feels heavy. There has been one high quality small 6-DOF device, the MPB Freedom 6S, on the market but it was very vulnerable and there were no compelling demos showing its potential.
grippers / end effectors
Not many user interfaces require a force feedback pen-style stylus, let alone a round knob, and it is certainly not easy to identify the "manipulandum" as they are called, with a 6-DOF object on the screen. A gripper would seem to be a more natural interface for a small haptics device, but good implementations are rare.