Game Development Reference
In-Depth Information
object is dictated directly by the operator, it in a way “escapes'' the direct control of
the simulation. This difference in state, caused by the difference between what the
operator does and what is actually allowed by the simulation, can lead to an instability
if the physical animation is not sturdy enough vis-à-vis this phenomenon. The essential
difficulty of haptic rendering (which differentiates it from other renderings) lies in the
bi-directional transfer of data, i.e. exchange of energy between the device and the
operator. In automation, this is reflected in a bilateral control type modelling or a
closed double loop.
The chapter on control of force feedback interfaces describes that haptic devices
operate on two different modes characterising the relation between the position (more
specifically, the speed) and the forces exerted by the device. In impedance mode, the
simulation receives the measured speed (actually, a succession of positions) of the
device. These positions determine the desired states of the avatar.
At each step of simulation, the animation engine corrects, if required, the state and
deduces the possible forces of interactions which it sends to the device (and which are
then exerted on the operator). This mode is used by most of the haptic devices, but it
does not provide an exact recreation of interactions with a rigid body (as the force sent
is bound to be limited and the interpenetrations cannot be avoided). The admittance
mode depends on a reverse exchange of data: The simulation recovers a force measured
by the device (which it considers as a part of external forces), deduces an acceleration,
a speed and then a position that the device has to then maintain. This control mode
is unstable in free zones (as the system has to decide its position as per the user's posi-
tion, which creates an artificial inertia detrimental to the quality of haptic rendering).
Moreover, a device in admittance mode requires force sensors, which increases its cost
and can reduce its sturdiness. We can observe that the data received from and provided
to the device vary in nature as per the mode selected. This means that the two modes
can be complementary to or dependent on the type of haptic interface used. In both
the cases, the data to be provided is calculated on the basis of the virtual environment.
However, the coupling of a haptic device with a virtual reality application imposes
some particularly severe constraints on the simulation, mainly in terms of haptic refresh
rate. In fact, a direct interface, i.e. a simple sampling of data from both sides, can gener-
ally make a system unstable after the oscillations caused during the interaction between
rigid objects. This problem was largely studied by a canonical example where a point
of a mass manipulated by the operator interacts with a virtual wall . In practice, the
force set point is updated at disconnected instants, which are not necessarily regu-
lar in time. A number of modules in simulation operate at variable times and certain
algorithms (for example, collision detection) are not time deterministic. Haptic infor-
mation is simply not defined between each simulation loop. This phenomenon related
to discrete time hampers an accurate processing of the contact when the operator is
manipulating the avatar. Figure 16.2 represents this entire phenomenon. When the
avatar is moving freely (i.e. when it is not in contact), the only forces transmitted to
the operator are inertia forces. When it interacts with other objects of the simulation,
the reaction forces calculated using various methods (penalties, impulses, constraints,
etc.) are predominant in the rendering. As a result of their rapid calculation, most of
the haptic rendering algorithms are based on penalty-based approaches.
In this case, higher the interpenetration between the avatar and the object con-
tacted, higher are the reaction forces. When these forces are too high, they make the
Search Nedrilad ::

Custom Search